diff --git a/web/src/engine/predictive-text/worker-thread/src/main/predict-helpers.ts b/web/src/engine/predictive-text/worker-thread/src/main/predict-helpers.ts
index 92691c8d20b..cd547cf6578 100644
--- a/web/src/engine/predictive-text/worker-thread/src/main/predict-helpers.ts
+++ b/web/src/engine/predictive-text/worker-thread/src/main/predict-helpers.ts
@@ -72,11 +72,30 @@ export const CORRECTION_SEARCH_THRESHOLDS = {
   REPLACEMENT_SEARCH_THRESHOLD: 4 as const // e^-4 = 0.0183156388.  Allows "80%" of an extra edit.
 }
 
+/**
+ * Tracks common intermediate prediction data, such as its underlying probabilities and its similarity to the actual context.
+ */
+export interface PredictionMetadata {
+  /**
+   * How directly the prediction matches the current token in the context.
+   *
+   * This is determined later in the suggestion-analysis project and is not
+   * available upon initial construction of this type.
+   */
+  matchLevel: SuggestionSimilarity;
+
+  /**
+   * Text from the triggering input that should _not_ be affected by the
+   * prediction.
+   */
+  preservationTransform: Transform;
+}
+
 /**
  * Collates information related to suggestions during the suggestion generation
  * process.
  */
-export type CorrectionPredictionTuple = {
+export interface CorrectionPredictionTupleCore {
   /**
    * The potential Suggestion (or Keep)
    */
@@ -90,19 +109,18 @@ export type CorrectionPredictionTuple = {
    * by the keystroke-sequence + correction likelihood.
    */
   totalProb: number;
+};
+
+export interface CorrectionPredictionTuple extends CorrectionPredictionTupleCore {
   /**
-   * How directly the prediction matches the current token in the context.
+   * Contains additional metadata about the prediction and its properties.
    *
-   * This is determined later in the suggestion-analysis project and is not
-   * available upon initial construction of this type.
+   * This object will generally remain unset by the `predictFromCorrections`
+   * method, with its values set afterward a layer or two removed from that
+   * specific call.
    */
-  matchLevel?: SuggestionSimilarity;
-  /**
-   * Text from the triggering input that should _not_ be affected by the
-   * prediction.
-   */
-  preservationTransform?: Transform;
-};
+  metadata: PredictionMetadata;
+}
 
 /**
  * An enum to be used when categorizing the level of similarity between
@@ -140,9 +158,15 @@ export enum SuggestionSimilarity {
   exact = 3
 }
 
-export function tupleDisplayOrderSort(a: CorrectionPredictionTuple, b: CorrectionPredictionTuple) {
+export function tupleDisplayOrderSort(
+  a: CorrectionPredictionTuple,
+  b: CorrectionPredictionTuple
+) {
+  const matchLevelA = a.metadata.matchLevel ?? 0;
+  const matchLevelB = b.metadata.matchLevel ?? 0;
+
   // Similarity distance
-  const simDist = (b.matchLevel ?? 0) - (a.matchLevel ?? 0);
+  const simDist = matchLevelB - matchLevelA;
   if(simDist != 0) {
     return simDist;
   }
@@ -177,7 +201,7 @@ export async function correctAndEnumerateWithoutTraversals(
   revertableTransitionId?: number
 }> {
   const inputTransform = transformDistribution[0].sample;
-  let rawPredictions: CorrectionPredictionTuple[] = [];
+  let rawPredictions: CorrectionPredictionTupleCore[] = [];
 
   let predictionRoots: ProbabilityMass<Transform>[];
 
@@ -212,13 +236,20 @@ export async function correctAndEnumerateWithoutTraversals(
 
   // Running in bulk over all suggestions, duplicate entries may be possible.
   rawPredictions = predictFromCorrections(lexicalModel, predictionRoots, context);
-  if(allowSpace) {
-    rawPredictions.forEach((entry) => entry.preservationTransform = inputTransform);
-  }
+  const predictions = rawPredictions.map((entry) => {
+    const preservationTransform = allowSpace ? inputTransform : null;
+    return {
+      ...entry,
+      metadata: {
+        preservationTransform,
+        matchLevel: SuggestionSimilarity.none // will be overwritten later
+      }
+    };
+  });
 
   return {
     postContextState: null,
-    rawPredictions: rawPredictions
+    rawPredictions: predictions
   };
 }
 
@@ -499,10 +530,17 @@ export function buildAndMapPredictions(
 
   // Worth considering:  extend Traversal to allow direct prediction lookups?
   // let traversal = match.finalTraversal; // ...
-  let predictions = predictFromCorrections(model, [predictionRoot], predictionContext);
-  predictions.forEach((entry) => {
-    entry.preservationTransform = tokenization.taillessTrueKeystroke;
+  let rawPredictions = predictFromCorrections(model, [predictionRoot], predictionContext);
+  const predictions = rawPredictions.map((entry) => {
     entry.prediction.sample.transform.deleteLeft += committedDeleteLeft;
+
+    return {
+      ...entry,
+      metadata: {
+        preservationTransform: tokenization.taillessTrueKeystroke,
+        matchLevel: SuggestionSimilarity.none // will be overwritten later
+      }
+    };
   });
 
   return predictions;
@@ -649,7 +687,7 @@ export async function correctAndEnumerate(
 export function shouldStopSearchingEarly(
   bestCorrectionCost: number,
   currentCorrectionCost: number,
-  rawPredictions: CorrectionPredictionTuple[]
+  rawPredictions: CorrectionPredictionTupleCore[]
 ) {
   if(currentCorrectionCost >= bestCorrectionCost + CORRECTION_SEARCH_THRESHOLDS.MAX_SEARCH_THRESHOLD) {
     return true;
@@ -659,10 +697,11 @@ export function shouldStopSearchingEarly(
       // Very useful for stopping 'sooner' when words reach a sufficient length.
       return true;
     } else {
-      // Sort the prediction list; we need them in descending order for the next check.
-      rawPredictions.sort(tupleDisplayOrderSort);
+      // Sort the prediction list; we need them in descending probability order
+      // for the next check.
+      rawPredictions.sort((a, b) => b.totalProb - a.totalProb);
 
-      // If the best suggestion from the search's current tier fails to beat the worst
+      // If the best result at the current state of the search fails to beat the worst
       // pending suggestion from previous tiers, assume all further corrections will
       // similarly fail to win; terminate the search-loop.
       if(rawPredictions[ModelCompositor.MAX_SUGGESTIONS-1].totalProb > Math.exp(-currentCorrectionCost)) {
@@ -690,8 +729,8 @@ export function predictFromCorrections(
   lexicalModel: LexicalModel,
   corrections: ProbabilityMass<Transform>[],
   context: Context
-): CorrectionPredictionTuple[] {
-  let returnedPredictions: CorrectionPredictionTuple[] = [];
+): CorrectionPredictionTupleCore[] {
+  let returnedPredictions: CorrectionPredictionTupleCore[] = [];
   const wordbreak = determineModelWordbreaker(lexicalModel);
 
   for(let correction of corrections) {
@@ -707,14 +746,13 @@ export function predictFromCorrections(
         pair.sample.transform.id = correctionTransform.id;
       }
 
-      let tuple: CorrectionPredictionTuple = {
+      let tuple: CorrectionPredictionTupleCore = {
         prediction: pair,
         correction: {
           sample: correctionRoot,
           p: correctionProb
         },
-        totalProb: pair.p * correctionProb,
-        matchLevel: SuggestionSimilarity.none
+        totalProb: pair.p * correctionProb
       };
       return tuple;
     });
@@ -847,7 +885,7 @@ export function processSimilarity(
     if(keyed(tuple.correction.sample) == keyedPrefix) {
       if(predictedWord == truePrefix) {
         // Exact match:  it's a perfect 'keep' suggestion.
-        tuple.matchLevel = SuggestionSimilarity.exact;
+        tuple.metadata.matchLevel = SuggestionSimilarity.exact;
         keepOption = toAnnotatedSuggestion(lexicalModel, tuple.prediction.sample, 'keep',  models.QuoteBehavior.noQuotes);
 
         // Indicates that this suggestion exists directly within the lexical
@@ -859,15 +897,15 @@ export function processSimilarity(
         keepOption = tuple.prediction.sample as Outcome<Keep>;
       } else if(keyCased(predictedWord) == lowercasedPrefix) {
         // Case-insensitive match.  No diacritic differences; the ONLY difference is casing.
-        tuple.matchLevel = SuggestionSimilarity.sameText;
+        tuple.metadata.matchLevel = SuggestionSimilarity.sameText;
       } else if(keyed(predictedWord) == keyedPrefix) {
         // Diacritic-insensitive / exact-key match.
-        tuple.matchLevel = SuggestionSimilarity.sameKey;
+        tuple.metadata.matchLevel = SuggestionSimilarity.sameKey;
       } else {
-        tuple.matchLevel = SuggestionSimilarity.none;
+        tuple.metadata.matchLevel = SuggestionSimilarity.none;
       }
     } else {
-      tuple.matchLevel = SuggestionSimilarity.none;
+      tuple.metadata.matchLevel = SuggestionSimilarity.none;
     }
   }
 
@@ -932,7 +970,10 @@ export function createDefaultKeep(
       sample: truePrefix,
       p: inputTransformProb * MAX_PROB
     },
-    matchLevel: SuggestionSimilarity.exact
+    metadata: {
+      preservationTransform: null,
+      matchLevel: SuggestionSimilarity.exact
+    }
   };
 }
 
@@ -1019,14 +1060,14 @@ export function predictionAutoSelect(suggestionDistribution: CorrectionPredictio
   // - such as replacing `cant` with `can't` if the latter is much more frequent -
   // we may wish to group matchLevel values below by 'mapping' them with an appropriate
   // function.  (Both on the next line and within the reduce functor.)
-  const bestSuggestionTier = bestSuggestion.matchLevel;
+  const bestSuggestionTier = bestSuggestion.metadata.matchLevel;
 
   // compare best vs other probabilities of compatible tier.
   const probSum = suggestionDistribution.reduce((accum, current) => {
     // If the suggestion is from a different similarity tier, do not count it against
     // the required auto-select probability ratio threshold.  That threshold should
     // only apply within the suggestion's tier.
-    return accum + (current.matchLevel == bestSuggestionTier ? current.totalProb : 0)
+    return accum + (current.metadata.matchLevel == bestSuggestionTier ? current.totalProb : 0)
   }, 0);
   const proportionOfBest = bestSuggestion.totalProb / probSum;
   if(proportionOfBest < AUTOSELECT_PROPORTION_THRESHOLD) {
@@ -1076,9 +1117,9 @@ export function finalizeSuggestions(
     // from prior tokens!
     //
     // Note:  may need adjustment if/when supporting phrase-level correction.
-    if(tuple.preservationTransform) {
-      const presDL = tuple.preservationTransform.deleteLeft;
-      const mergedTransform = models.buildMergedTransform(tuple.preservationTransform, prediction.sample.transform);
+    if(tuple.metadata.preservationTransform) {
+      const presDL = tuple.metadata.preservationTransform.deleteLeft;
+      const mergedTransform = models.buildMergedTransform(tuple.metadata.preservationTransform, prediction.sample.transform);
       // Any preserved delete-left is applied early because it directly affects the suggestion
       // root; we need to remove that preserved delete-left here.
       if(presDL > 0) {
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/correction-search/early-correction-search-stopping.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/correction-search/early-correction-search-stopping.tests.ts
index bf45e3b94e1..2afd5f5290a 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/correction-search/early-correction-search-stopping.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/correction-search/early-correction-search-stopping.tests.ts
@@ -1,6 +1,6 @@
 import { assert } from 'chai';
 
-import { CORRECTION_SEARCH_THRESHOLDS, CorrectionPredictionTuple, ModelCompositor, shouldStopSearchingEarly } from "@keymanapp/lm-worker/test-index";
+import { CORRECTION_SEARCH_THRESHOLDS, CorrectionPredictionTupleCore, ModelCompositor, shouldStopSearchingEarly } from "@keymanapp/lm-worker/test-index";
 
 describe('correction-search: shouldStopSearchingEarly', () => {
   it('stops early once new corrections are less likely than currently discovered predictions', () => {
@@ -16,7 +16,7 @@ describe('correction-search: shouldStopSearchingEarly', () => {
     const predictions = predictionProbs.map((entry) => {
       return {
         totalProb: entry
-      } as CorrectionPredictionTuple
+      } as CorrectionPredictionTupleCore
     });
 
     // Thresholding is performed in log-space.
@@ -33,8 +33,8 @@ describe('correction-search: shouldStopSearchingEarly', () => {
     //
     // Can technically run the method with an empty array, but the actual scenario would have
     // at least one prediction present in the "found predictions" array.
-    assert.isFalse(shouldStopSearchingEarly(baseCost, baseCost + expectedThreshold - 0.01, [{ totalProb: Math.exp(-1) } as CorrectionPredictionTuple]));
-    assert.isTrue(shouldStopSearchingEarly( baseCost, baseCost + expectedThreshold + 0.01, [{ totalProb: Math.exp(-1) } as CorrectionPredictionTuple]));
+    assert.isFalse(shouldStopSearchingEarly(baseCost, baseCost + expectedThreshold - 0.01, [{ totalProb: Math.exp(-1) } as CorrectionPredictionTupleCore]));
+    assert.isTrue(shouldStopSearchingEarly( baseCost, baseCost + expectedThreshold + 0.01, [{ totalProb: Math.exp(-1) } as CorrectionPredictionTupleCore]));
   });
 
   it('stops checking corrections earlier when enough predictions have been found', () => {
@@ -46,7 +46,7 @@ describe('correction-search: shouldStopSearchingEarly', () => {
     const predictions = predictionProbs.map((entry) => {
       return {
         totalProb: entry
-      } as CorrectionPredictionTuple
+      } as CorrectionPredictionTupleCore
     });
 
     const baseCost = 1;
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/auto-correct.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/auto-correct.tests.ts
index 882390591a9..354324210ef 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/auto-correct.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/auto-correct.tests.ts
@@ -1,6 +1,19 @@
 import { assert } from 'chai';
 
-import { AUTOSELECT_PROPORTION_THRESHOLD, CorrectionPredictionTuple, predictionAutoSelect, SuggestionSimilarity, tupleDisplayOrderSort } from "@keymanapp/lm-worker/test-index";
+import {
+  AUTOSELECT_PROPORTION_THRESHOLD,
+  CorrectionPredictionTuple,
+  predictionAutoSelect,
+  PredictionMetadata,
+  SuggestionSimilarity,
+  tupleDisplayOrderSort
+} from "@keymanapp/lm-worker/test-index";
+
+const defaultMetadata: PredictionMetadata = {
+  matchLevel: SuggestionSimilarity.none,
+  preservationTransform: undefined
+}
+
 /*
   * Preconditions:
   * - there should always be a 'keep' option.  Now, whether or not that option
@@ -35,7 +48,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 1
         },
-        totalProb: 1
+        totalProb: 1,
+        metadata: defaultMetadata
       }
     ];
 
@@ -66,7 +80,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 0.01
         },
-        totalProb: 0.01
+        totalProb: 0.01,
+        metadata: defaultMetadata
       },
       {
         correction: {
@@ -84,7 +99,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 0.8
         },
-        totalProb: 0.8
+        totalProb: 0.8,
+        metadata: defaultMetadata
       }
     ];
 
@@ -115,7 +131,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 1
         },
-        totalProb: 1
+        totalProb: 1,
+        metadata: defaultMetadata
       }
     ];
 
@@ -145,10 +162,11 @@ describe('predictionAutoSelect', () => {
         },
         p: .05
       },
-      totalProb: .04
+      totalProb: .04,
+      metadata: defaultMetadata
     }
 
-    const highestNonKeepSuggestion: CorrectionPredictionTuple = {
+    const highestNonKeepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -163,7 +181,8 @@ describe('predictionAutoSelect', () => {
         },
         p: .55
       },
-      totalProb: .44
+      totalProb: .44,
+      metadata: defaultMetadata
     };
 
     const predictions: CorrectionPredictionTuple[] = [
@@ -184,7 +203,8 @@ describe('predictionAutoSelect', () => {
           },
           p: .4
         },
-        totalProb: .32
+        totalProb: .32,
+        metadata: defaultMetadata
       },
       {
         correction: {
@@ -201,7 +221,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 1
         },
-        totalProb: .2
+        totalProb: .2,
+        metadata: defaultMetadata
       }
     ];
 
@@ -214,7 +235,7 @@ describe('predictionAutoSelect', () => {
   });
 
   it(`selects solitary non-'keep' suggestion when 'keep' does not match model`, () => {
-    const keepSuggestion: CorrectionPredictionTuple = {
+    const keepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -231,14 +252,15 @@ describe('predictionAutoSelect', () => {
         },
         p: .05
       },
-      totalProb: .04
+      totalProb: .04,
+      metadata: defaultMetadata
     }
 
     // To 'win', a suggestion (currently) needs at least twice the probability of the sum of all alternatives.
     // This threshold may be subject to change.
     //
     // Refer to AUTOSELECT_PROPORTION_THRESHOLD in predict-helpers.ts.
-    const onlyNonKeepSuggestion: CorrectionPredictionTuple = {
+    const onlyNonKeepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -253,7 +275,8 @@ describe('predictionAutoSelect', () => {
         },
         p: .01
       },
-      totalProb: .008
+      totalProb: .008,
+      metadata: defaultMetadata
     };
 
     const predictions: CorrectionPredictionTuple[] = [
@@ -275,7 +298,7 @@ describe('predictionAutoSelect', () => {
   });
 
   it(`does not select non-'keep' without sufficient winning probability`, () => {
-    const keepSuggestion: CorrectionPredictionTuple = {
+    const keepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -292,14 +315,15 @@ describe('predictionAutoSelect', () => {
         },
         p: .05
       },
-      totalProb: .04
+      totalProb: .04,
+      metadata: defaultMetadata
     }
 
     // To 'win', a suggestion (currently) needs at least twice the probability of the sum of all alternatives.
     // This threshold may be subject to change.
     //
     // Refer to AUTOSELECT_PROPORTION_THRESHOLD in predict-helpers.ts.
-    const highestNonKeepSuggestion: CorrectionPredictionTuple = {
+    const highestNonKeepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -314,7 +338,8 @@ describe('predictionAutoSelect', () => {
         },
         p: .55
       },
-      totalProb: .44
+      totalProb: .44,
+      metadata: defaultMetadata
     };
 
     const predictions: CorrectionPredictionTuple[] = [
@@ -335,7 +360,8 @@ describe('predictionAutoSelect', () => {
           },
           p: .4
         },
-        totalProb: .32
+        totalProb: .32,
+        metadata: defaultMetadata
       },
       {
         correction: {
@@ -352,7 +378,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 1
         },
-        totalProb: .2
+        totalProb: .2,
+        metadata: defaultMetadata
       }
     ];
 
@@ -370,7 +397,7 @@ describe('predictionAutoSelect', () => {
   });
 
   it(`does select non-'keep' with sufficient winning probability`, () => {
-    const keepSuggestion: CorrectionPredictionTuple = {
+    const keepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .8
@@ -387,10 +414,11 @@ describe('predictionAutoSelect', () => {
         },
         p: .05
       },
-      totalProb: .04
+      totalProb: .04,
+      metadata: defaultMetadata
     }
 
-    const highestNonKeepSuggestion: CorrectionPredictionTuple = {
+    const highestNonKeepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thin',
         p: .9
@@ -405,7 +433,8 @@ describe('predictionAutoSelect', () => {
         },
         p: .75
       },
-      totalProb: .675
+      totalProb: .675,
+      metadata: defaultMetadata
     };
 
     const predictions: CorrectionPredictionTuple[] = [
@@ -426,7 +455,8 @@ describe('predictionAutoSelect', () => {
           },
           p: .2
         },
-        totalProb: .18
+        totalProb: .18,
+        metadata: defaultMetadata
       },
       {
         correction: {
@@ -443,7 +473,8 @@ describe('predictionAutoSelect', () => {
           },
           p: 1
         },
-        totalProb: .1
+        totalProb: .1,
+        metadata: defaultMetadata
       }
     ];
 
@@ -459,7 +490,7 @@ describe('predictionAutoSelect', () => {
   });
 
   it('ignores non key-matched suggestions when key-matched suggestions exist', () => {
-    const keepSuggestion: CorrectionPredictionTuple = {
+    const keepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'cant',
         p: 1
@@ -477,10 +508,10 @@ describe('predictionAutoSelect', () => {
         p: 1
       },
       totalProb: 1,
-      matchLevel: SuggestionSimilarity.exact
+      metadata: { matchLevel: SuggestionSimilarity.exact, preservationTransform: null }
     }
 
-    const expectedSuggestion: CorrectionPredictionTuple = {
+    const expectedSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'cant',
         p: 1
@@ -496,7 +527,7 @@ describe('predictionAutoSelect', () => {
         p: .2
       },
       totalProb: .2,
-      matchLevel: SuggestionSimilarity.sameKey
+      metadata: { matchLevel: SuggestionSimilarity.sameKey, preservationTransform: null }
     };
 
     const predictions: CorrectionPredictionTuple[] = [
@@ -518,7 +549,7 @@ describe('predictionAutoSelect', () => {
           p: .8
         },
         totalProb: .8,
-        matchLevel: SuggestionSimilarity.none
+        metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: null }
       }
     ];
 
@@ -534,7 +565,7 @@ describe('predictionAutoSelect', () => {
   // The idea:  avoid "over-correcting" when a potential correction has a
   // super-high-frequency word.
   it('does not auto-select suggestion if its root correction is not most likely', () => {
-    const keepSuggestion: CorrectionPredictionTuple = {
+    const keepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thi',
         p: .7
@@ -551,10 +582,11 @@ describe('predictionAutoSelect', () => {
         },
         p: .05
       },
-      totalProb: .035
+      totalProb: .035,
+      metadata: defaultMetadata
     }
 
-    const highestCorrectionSuggestion: CorrectionPredictionTuple = {
+    const highestCorrectionSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'thi',
         p: .7
@@ -569,10 +601,11 @@ describe('predictionAutoSelect', () => {
         },
         p: .1
       },
-      totalProb: .07
+      totalProb: .07,
+      metadata: defaultMetadata
     };
 
-    const highestNonKeepSuggestion: CorrectionPredictionTuple = {
+    const highestNonKeepSuggestion: CorrectionPredictionTuple= {
       correction: {
         sample: 'the',
         p: .3
@@ -587,7 +620,8 @@ describe('predictionAutoSelect', () => {
         },
         p: 1
       },
-      totalProb: .3
+      totalProb: .3,
+      metadata: defaultMetadata
     };
 
     const predictions: CorrectionPredictionTuple[] = [
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/build-and-map-predictions.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/build-and-map-predictions.tests.ts
index 73d8276e132..a2729d36334 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/build-and-map-predictions.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/build-and-map-predictions.tests.ts
@@ -46,7 +46,6 @@ describe('buildAndMapPredictions', () => {
     ];
 
     const basePredictions = predictFromCorrections(plainModel, correctionDistribution, context);
-    basePredictions.forEach((entry) => assert.isNotOk(entry.preservationTransform));
 
     // must construct the taillessTrueKeystroke appropriately.
     const tailless = { insert: 'TEST', deleteLeft: 0 };
@@ -64,8 +63,8 @@ describe('buildAndMapPredictions', () => {
     );
 
     assert.deepEqual(mappedPredictions.map((tuple) => tuple.prediction), basePredictions.map((tuple) => tuple.prediction));
-    mappedPredictions.forEach((tuple) => assert.isOk(tuple.preservationTransform));
-    mappedPredictions.forEach((tuple) => tuple.preservationTransform == tailless);
+    mappedPredictions.forEach((tuple) => assert.isOk(tuple.metadata?.preservationTransform));
+    mappedPredictions.forEach((tuple) => tuple.metadata?.preservationTransform == tailless);
   });
 
   it('properly handles empty prediction roots from deleted same-token codepoints', () => {
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/create-default-keep.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/create-default-keep.tests.ts
index 6b5981c8343..3c093a1af99 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/create-default-keep.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/create-default-keep.tests.ts
@@ -12,7 +12,12 @@ import { assert } from 'chai';
 import { LexicalModelTypes } from "@keymanapp/common-types";
 import * as wordBreakers from '@keymanapp/models-wordbreakers';
 
-import { CorrectionPredictionTuple, createDefaultKeep, models, SuggestionSimilarity } from "@keymanapp/lm-worker/test-index";
+import {
+  CorrectionPredictionTuple,
+  createDefaultKeep,
+  models,
+  SuggestionSimilarity
+} from "@keymanapp/lm-worker/test-index";
 
 import CasingFunction = LexicalModelTypes.CasingFunction;
 import Context = LexicalModelTypes.Context;
@@ -126,7 +131,7 @@ describe('produceKeep', () => {
         p: 1
       },
       totalProb: 1,
-      matchLevel: SuggestionSimilarity.exact
+      metadata: { matchLevel: SuggestionSimilarity.exact, preservationTransform: null }
     };
 
     const tuple = createDefaultKeep(testModelWithCasing, context, trueInput);
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/predict-from-corrections.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/predict-from-corrections.tests.ts
index 31f5063c15c..d18f2878418 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/predict-from-corrections.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/predict-from-corrections.tests.ts
@@ -4,7 +4,7 @@ import { assert } from 'chai';
 import { deepCopy } from "keyman/common/web-utils";
 import { LexicalModelTypes } from '@keymanapp/common-types';
 
-import { models, predictFromCorrections, tupleDisplayOrderSort } from "@keymanapp/lm-worker/test-index";
+import { models, predictFromCorrections } from "@keymanapp/lm-worker/test-index";
 
 import CasingFunction = LexicalModelTypes.CasingFunction;
 import Context = LexicalModelTypes.Context;
@@ -115,7 +115,7 @@ describe('predictFromCorrections', () => {
     const predictions = predictFromCorrections(model, correctionDistribution, context);
     predictions.forEach((entry) => assert.equal(entry.correction.sample, 'Its'));
     predictions.forEach((entry) => assert.equal(entry.correction.p, 0.6));
-    predictions.sort(tupleDisplayOrderSort);
+    predictions.sort((a, b) => b.totalProb - a.totalProb);
 
     assert.sameDeepOrderedMembers(predictions.map((entry) => entry.prediction.sample), dummied_suggestions);
 
@@ -167,7 +167,7 @@ describe('predictFromCorrections', () => {
     const predictions = predictFromCorrections(model, correctionDistribution, context);
     predictions.forEach((entry) => assert.equal(entry.correction.sample, 'Its'));
     predictions.forEach((entry) => assert.equal(entry.correction.p, 0.6));
-    predictions.sort(tupleDisplayOrderSort);
+    predictions.sort((a, b) => b.totalProb - a.totalProb);
 
     assert.sameOrderedMembers(predictions.map((entry) => entry.prediction.sample.displayAs), ["it's", "its"]);
     assert.sameDeepOrderedMembers(predictions.map((entry) => entry.prediction.sample), dummied_suggestions.map((entry) => {
@@ -247,7 +247,7 @@ describe('predictFromCorrections', () => {
     });
 
     const predictions = predictFromCorrections(model, correctionDistribution, context);
-    predictions.sort(tupleDisplayOrderSort);
+    predictions.sort((a, b) => b.totalProb - a.totalProb);
 
     assert.sameOrderedMembers(predictions.map((entry) => entry.prediction.sample.displayAs), ["is", "it's", "isn't", "its"]);
     assert.sameDeepMembers(predictions.map((entry) => entry.prediction.sample), dummied_suggestions.flatMap((entry) => entry));
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-deduplication.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-deduplication.tests.ts
index bfd2eeefafc..6a78ef1fd8e 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-deduplication.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-deduplication.tests.ts
@@ -4,7 +4,13 @@ import * as wordBreakers from '@keymanapp/models-wordbreakers';
 import { deepCopy } from 'keyman/common/web-utils';
 import { LexicalModelTypes } from '@keymanapp/common-types';
 
-import { CorrectionPredictionTuple, dedupeSuggestions, models } from "@keymanapp/lm-worker/test-index";
+import {
+  CorrectionPredictionTuple,
+  CorrectionPredictionTupleCore,
+  SuggestionSimilarity,
+  dedupeSuggestions,
+  models
+} from "@keymanapp/lm-worker/test-index";
 
 import Context = LexicalModelTypes.Context;
 import DummyModel = models.DummyModel;
@@ -18,13 +24,23 @@ const testModel = new DummyModel({
   // No suggestions needed here, so we don't define any.
 });
 
+const mockMetadata: (tc: CorrectionPredictionTupleCore) => CorrectionPredictionTuple = (t: CorrectionPredictionTupleCore) => {
+  return {
+    ...t,
+    metadata: {
+      preservationTransform: null,
+      matchLevel: SuggestionSimilarity.none
+    }
+  }
+};
+
 /**
  * Builds a fresh copy of test values useful for suggestion-similarity
  * testing.
  * @returns
  */
 const build_its_is_set = () => {
-  const its: CorrectionPredictionTuple = {
+  const its: CorrectionPredictionTupleCore = {
     correction: {
       sample: 'its',
       p: 0.8
@@ -43,7 +59,7 @@ const build_its_is_set = () => {
     // matchLevel does not yet exist.
   };
 
-  const it_is: CorrectionPredictionTuple = {
+  const it_is: CorrectionPredictionTupleCore = {
     correction: {
       sample: 'its',
       p: 0.8
@@ -61,7 +77,7 @@ const build_its_is_set = () => {
     totalProb: 0.64
   };
 
-  const is: CorrectionPredictionTuple = {
+  const is: CorrectionPredictionTupleCore = {
     correction: {
       sample: 'is',
       p: 0.2
@@ -79,7 +95,7 @@ const build_its_is_set = () => {
     totalProb: 0.1
   };
 
-  const is_not: CorrectionPredictionTuple = {
+  const is_not: CorrectionPredictionTupleCore = {
     correction: {
       sample: 'is',
       p: 0.2
@@ -115,7 +131,7 @@ describe('dedupeSuggestions', () => {
     };
 
     const testSet = build_its_is_set();
-    const predictions = [...Object.values(testSet)];
+    const predictions: CorrectionPredictionTuple[] = [...Object.values(testSet)].map(mockMetadata) ;
 
     const deduplicated = dedupeSuggestions(testModel, predictions, context);
 
@@ -136,10 +152,10 @@ describe('dedupeSuggestions', () => {
       ...Object.values(testSet).map((entry) => deepCopy(entry)),
       ...Object.values(testSet).map((entry) => deepCopy(entry)),
       deepCopy(testSet.it_is) // as in, `it's`, the contraction.
-    ];
+    ].map(mockMetadata);
 
     const deduplicated = dedupeSuggestions(testModel, predictions, context);
-    const expected = [...Object.values(testSet)];
+    const expected = [...Object.values(testSet)].map(mockMetadata);
     // Note:  only changes the _total_ probability.
     //
     // There's no mathematically safe way to combine the components if the
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-finalization.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-finalization.tests.ts
index 387bf6dab93..01b017e5fee 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-finalization.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-finalization.tests.ts
@@ -5,7 +5,7 @@ import { deepCopy } from 'keyman/common/web-utils';
 import * as wordBreakers from '@keymanapp/models-wordbreakers';
 import { LexicalModelTypes } from '@keymanapp/common-types';
 
-import { CorrectionPredictionTuple, finalizeSuggestions, models } from "@keymanapp/lm-worker/test-index";
+import { CorrectionPredictionTuple, PredictionMetadata, SuggestionSimilarity, finalizeSuggestions, models } from "@keymanapp/lm-worker/test-index";
 
 import DummyModel = models.DummyModel;
 import Outcome = LexicalModelTypes.Outcome;
@@ -48,6 +48,11 @@ const testModelWithoutSpacing = new DummyModel({
 const build_its_is_set = (verbose?: string) => {
   const verboseFlag = (verbose == 'verbose' ? true : false);
 
+  const metadata: PredictionMetadata = {
+    matchLevel: SuggestionSimilarity.none,
+    preservationTransform: undefined
+  };
+
   const its: CorrectionPredictionTuple = {
     correction: {
       sample: 'its',
@@ -63,8 +68,8 @@ const build_its_is_set = (verbose?: string) => {
       },
       p: 0.2
     },
-    totalProb: 0.16
-    // matchLevel does not yet exist.
+    totalProb: 0.16,
+    metadata: {...metadata}
   };
 
   const it_is: CorrectionPredictionTuple = {
@@ -82,7 +87,8 @@ const build_its_is_set = (verbose?: string) => {
       },
       p: 0.8
     },
-    totalProb: 0.64
+    totalProb: 0.64,
+    metadata: {...metadata}
   };
 
   const is: CorrectionPredictionTuple = {
@@ -100,7 +106,8 @@ const build_its_is_set = (verbose?: string) => {
       },
       p: 0.5
     },
-    totalProb: 0.1
+    totalProb: 0.1,
+    metadata: {...metadata}
   };
 
   const is_not: CorrectionPredictionTuple = {
@@ -118,7 +125,8 @@ const build_its_is_set = (verbose?: string) => {
       },
       p: 0.5
     },
-    totalProb: 0.1
+    totalProb: 0.1,
+    metadata: {...metadata}
   };
 
   const baseDefinitions = {
diff --git a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-similarity.tests.ts b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-similarity.tests.ts
index 6cfa731429d..9672b102dff 100644
--- a/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-similarity.tests.ts
+++ b/web/src/test/auto/headless/engine/predictive-text/worker-thread/prediction-helpers/suggestion-similarity.tests.ts
@@ -5,7 +5,14 @@ import * as wordBreakers from '@keymanapp/models-wordbreakers';
 import { deepCopy } from 'keyman/common/web-utils';
 import { LexicalModelTypes } from '@keymanapp/common-types';
 
-import { CorrectionPredictionTuple, models, processSimilarity, SuggestionSimilarity, toAnnotatedSuggestion } from "@keymanapp/lm-worker/test-index";
+import {
+  CorrectionPredictionTuple,
+  models,
+  PredictionMetadata,
+  processSimilarity,
+  SuggestionSimilarity,
+  toAnnotatedSuggestion
+} from "@keymanapp/lm-worker/test-index";
 
 import CasingFunction = LexicalModelTypes.CasingFunction;
 import Context = LexicalModelTypes.Context;
@@ -109,6 +116,11 @@ const testModelWithCasing = new DummyModel({
  * @returns
  */
 const build_its_is_set = () => {
+  const metadata: PredictionMetadata = {
+    matchLevel: SuggestionSimilarity.none,
+    preservationTransform: undefined
+  };
+
   const its: CorrectionPredictionTuple = {
     correction: {
       sample: 'its',
@@ -124,8 +136,8 @@ const build_its_is_set = () => {
       },
       p: 0.2
     },
-    totalProb: 0.16
-    // matchLevel does not yet exist.
+    totalProb: 0.16,
+    metadata: {...metadata}
   };
 
   const it_is: CorrectionPredictionTuple = {
@@ -143,7 +155,8 @@ const build_its_is_set = () => {
       },
       p: 0.8
     },
-    totalProb: 0.64
+    totalProb: 0.64,
+    metadata: {...metadata}
   };
 
   const is: CorrectionPredictionTuple = {
@@ -161,7 +174,8 @@ const build_its_is_set = () => {
       },
       p: 0.5
     },
-    totalProb: 0.1
+    totalProb: 0.1,
+    metadata: {...metadata}
   };
 
   const is_not: CorrectionPredictionTuple = {
@@ -179,7 +193,8 @@ const build_its_is_set = () => {
       },
       p: 0.5
     },
-    totalProb: 0.1
+    totalProb: 0.1,
+    metadata: {...metadata}
   };
 
   return {
@@ -213,16 +228,16 @@ describe('processSimilarity', () => {
     const expectation: CorrectionPredictionTuple[] = [
       {
         ...testSet.its,
-        matchLevel: SuggestionSimilarity.exact
+        metadata: { matchLevel: SuggestionSimilarity.exact, preservationTransform: undefined }
       }, {
         ...testSet.it_is,
-        matchLevel: SuggestionSimilarity.sameKey
+        metadata: { matchLevel: SuggestionSimilarity.sameKey, preservationTransform: undefined }
       }, {
         ...testSet.is,
-        matchLevel: SuggestionSimilarity.none
+        metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
       }, {
         ...testSet.is_not,
-        matchLevel: SuggestionSimilarity.none
+        metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
       }
     ];
 
@@ -260,16 +275,16 @@ describe('processSimilarity', () => {
     const expectation: CorrectionPredictionTuple[] = [
       {
         ...testSet.its,
-        matchLevel: SuggestionSimilarity.sameKey
+        metadata: { matchLevel: SuggestionSimilarity.sameKey, preservationTransform: undefined }
       }, {
         ...testSet.it_is,
-        matchLevel: SuggestionSimilarity.exact
+        metadata: { matchLevel: SuggestionSimilarity.exact, preservationTransform: undefined }
       }, {
         ...testSet.is,
-        matchLevel: SuggestionSimilarity.none
+        metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
       }, {
         ...testSet.is_not,
-        matchLevel: SuggestionSimilarity.none
+        metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
       }
     ];
 
@@ -324,17 +339,17 @@ describe('processSimilarity', () => {
       const expectation: CorrectionPredictionTuple[] = [
         {
           ...testSet.its,
-          matchLevel: SuggestionSimilarity.sameKey
+          metadata: { matchLevel: SuggestionSimilarity.sameKey, preservationTransform: undefined }
         }, {
           ...testSet.it_is,
           // case mismatch, detectable because we have access to a lowercasing/uppercasing function.
-          matchLevel: SuggestionSimilarity.sameText
+          metadata: { matchLevel: SuggestionSimilarity.sameText, preservationTransform: undefined }
         }, {
           ...testSet.is,
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }, {
           ...testSet.is_not,
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }
       ];
 
@@ -378,17 +393,17 @@ describe('processSimilarity', () => {
       const expectation: CorrectionPredictionTuple[] = [
         {
           ...testSet.its,
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }, {
           ...testSet.it_is,
           // case mismatch, detectable because we have access to a lowercasing/uppercasing function.
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }, {
           ...testSet.is,
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }, {
           ...testSet.is_not,
-          matchLevel: SuggestionSimilarity.none
+          metadata: { matchLevel: SuggestionSimilarity.none, preservationTransform: undefined }
         }
       ];