Release 1.11.0 (Release C): GIT-86, 98, 99, 101, 102, 73, 105, 100, 107, 109, 83, 112

[nTEG-dev]

What

Release 1.11.0 (Release C): 11 PRs stacked on origin/main (b1b3cb9, 1.10.0), then 7a3ac5c bumping the version to 1.11.0. CHANGELOG [1.11.0] - 2026-09-21, with a fresh [Unreleased] above it. Not merged, tagged or published yet. Phase 2 waits for Chris's go.

Merge order: #39, #37, #38, #40, #41, #42, #44, #43, #45, #46, #47.

PR	Ticket
#39	GIT-99, GIT-112	hooks resolve .gitmem like the server; Stop enforcement without jq; payload path named
#37	GIT-86	never resume a session of another project
#38	GIT-98	per-row delta sync of the embedding index
#40	GIT-101	every write tool reports durable / stored_in
#41	GIT-102	bounded write-path check; self-describing close WARNs
#42	GIT-73	metrics wait for their session row
#44	GIT-105	thread triples reference the thread row id
#43	GIT-100	migrate-root merges memory files by record id
#45	GIT-107	roots compared by real path
#46	GIT-109	static schema-drift gate against v1.8.0
#47	GIT-83	release gate; publish depends on the venue and hook jobs

Conflict resolutions

All were of the kinds agreed in advance. Nothing else came up.

1. Test mocks missing findSessionById (fix: never resume a session of another project (GIT-86) #37), in 4 files, not just the one expected: session-close-validation, session-close-payload-path (fix: hooks resolve .gitmem like the server; session_close names the payload path (GIT-99) #39 copied its mock header from it), session-close-supabase and session-close-bridge. Each gets findSessionById → null (abb772a). Nothing failed before the fix, because no test reaches that fallback.
2. blank-supabase.mjs: conflicts at fix: sync only changed rows on a learnings-fingerprint miss (GIT-98) #38, fix: every write tool reports durability; resolve_thread stops claiming success (GIT-101) #40, fix: bounded write-path check with an UNVERIFIED verdict; self-describing close WARNs (GIT-102) #41, fix: thread triples reference the thread row id, not its t- id (GIT-105) #44 and fix: compare .gitmem roots by real path, not string (GIT-107) #45. Every mode, step, result field and check is kept. Order: fix: every write tool reports durability; resolve_thread stops claiming success (GIT-101) #40's resolve step, then fix: hooks resolve .gitmem like the server; session_close names the payload path (GIT-99) #39's no-payload close, then the real close. EXPECTED_FAILURES: fix: metrics wait for their session row instead of racing it to a 409 (GIT-73) #42 removed GIT-73 and fix: thread triples reference the thread row id, not its t- id (GIT-105) #44 removed GIT-105; both removals are kept, so the list is empty.
3. session-close.ts sessionRowStored:
   • at fix: every write tool reports durability; resolve_thread stops claiming success (GIT-101) #40: kept alongside fix: hooks resolve .gitmem like the server; session_close names the payload path (GIT-99) #39's payloadReadError (separate declarations);
   • at fix: bounded write-path check with an UNVERIFIED verdict; self-describing close WARNs (GIT-102) #41: fix: every write tool reports durability; resolve_thread stops claiming success (GIT-101) #40 and fix: bounded write-path check with an UNVERIFIED verdict; self-describing close WARNs (GIT-102) #41 add the same flag, so one declaration and one assignment (after the upsert) are kept, with one comment covering both uses;
   • fix: metrics wait for their session row instead of racing it to a 409 (GIT-73) #42 doesn't touch this file.
4. tsc, a return without durable / stored_in, at fix: every write tool reports durability; resolve_thread stops claiming success (GIT-101) #40: fix: hooks resolve .gitmem like the server; session_close names the payload path (GIT-99) #39's "closing-payload.json not found" return predates WriteResult. It's a pre-write exit, so it gets ...notStored(). That was the only one; tsc was clean after every other merge.

Gates (on this branch)

Gate	Result
tsc --noEmit	✅ clean, run after every merge
Unit (vitest run)	✅ 1357 / 1357 (94 files)
Hooks suite, emulated slim PATH (no jq, no ps, env -i)	✅ 46 / 46
Venue vujndhcumgroztfkkdus reset to v1.8.0:schema/setup.sql → flow	✅ exit 0: close persisted, relevance readable, threads 3 = 3, resolve durable, no-payload close names the path, self-describing WARNs, health write-path verdict, 3 thread triples linked, no false stranded notice
→ egress --rows 250 --usage 5	✅ exit 0; no full re-download after the cold start (921,750 B cold, then 100 B / 42,823 B)
→ projects	✅ exit 0 (11/11 cross-project checks)
current setup.sql applied twice → flow	✅ both applications clean; exit 0
EXPECTED_FAILURES	✅ empty; 0 known kinds seen in all four runs. The only non-2xx is the read-only capability probe for task_observations, which the driver treats as expected.
Drift analyzer	✅ the same 11 known findings (3 GIT-109 probe-guarded, 8 GIT-106 dev tier), nothing new; 77 call sites, 384 accesses, 5 known unresolved

The venue was restored to v1.8.0 afterwards. Production (cjptxyezuxdiinufgrrm) was never touched.

CI gate on GitHub

✅ Run 35687261565 (workflow_dispatch on release/C at f3fdae3): build (18, 20, 22) ✅, hooks-without-jq ✅, venue-gate ✅. Every venue step ran, none were skipped: the v1.8.0 reset via psql, then flow, egress and projects, then the current setup.sql applied twice, then flow. publish was skipped, as it should be (not a tag).

Release

Version 1.11.0. Phase 2, after Chris says go: preflight, merge commit, main's CI, tag v1.11.0, tag CI through both gates and publish, registry verification, GitHub Release.

🤖 Generated with Claude Code