Problem
Trigger.dev's realtime React hooks currently pull in @electric-sql/client@1.0.14 through @trigger.dev/core. In production we are seeing repeated browser-side unhandled rejections while using useRealtimeRun and useRealtimeBatch:
AbortError: signal is aborted without reason
at PrefetchQueue.abort
@electric-sql/client@1.0.14/dist/index.mjs
Related events are also reported as:
UnhandledRejection: Non-Error promise rejection captured with value: pause-stream
Across the related Sentry groupings we have observed 314 events affecting 7 users. The errors occur around Trigger.dev realtime requests and tab/subscription lifecycle changes.
Root cause
This appears to be ElectricSQL's stale aborted-prefetch bug:
ElectricSQL describes the failure mode as aborted prefetch promises remaining in the queue and being returned when a stream resumes, causing an AbortError to propagate and stop syncing. The fix clears the prefetch queue after abort and avoids returning aborted requests. It was released starting with @electric-sql/client@1.3.0.
Trigger.dev 4.5.4 still pins @electric-sql/client to 1.0.14 in packages/core/package.json.
Impact
Our application currently has a separate status-polling fallback, so captured generations generally recover. We are moving toward realtime-only status delivery; without the fallback, a stopped stream can leave completed runs stuck in a non-terminal UI state.
We would prefer not to override or patch Trigger.dev's transitive dependency in the application.
Request
Please update @electric-sql/client to a version containing the upstream fix and publish it through a supported Trigger.dev release.
A regression test covering tab hide/show or stream pause/resume with an in-flight prefetched request would also help ensure useRealtimeRun and useRealtimeBatch recover reliably.
Problem
Trigger.dev's realtime React hooks currently pull in
@electric-sql/client@1.0.14through@trigger.dev/core. In production we are seeing repeated browser-side unhandled rejections while usinguseRealtimeRunanduseRealtimeBatch:Related events are also reported as:
Across the related Sentry groupings we have observed 314 events affecting 7 users. The errors occur around Trigger.dev realtime requests and tab/subscription lifecycle changes.
Root cause
This appears to be ElectricSQL's stale aborted-prefetch bug:
ElectricSQL describes the failure mode as aborted prefetch promises remaining in the queue and being returned when a stream resumes, causing an
AbortErrorto propagate and stop syncing. The fix clears the prefetch queue after abort and avoids returning aborted requests. It was released starting with@electric-sql/client@1.3.0.Trigger.dev 4.5.4 still pins
@electric-sql/clientto1.0.14inpackages/core/package.json.Impact
Our application currently has a separate status-polling fallback, so captured generations generally recover. We are moving toward realtime-only status delivery; without the fallback, a stopped stream can leave completed runs stuck in a non-terminal UI state.
We would prefer not to override or patch Trigger.dev's transitive dependency in the application.
Request
Please update
@electric-sql/clientto a version containing the upstream fix and publish it through a supported Trigger.dev release.A regression test covering tab hide/show or stream pause/resume with an in-flight prefetched request would also help ensure
useRealtimeRunanduseRealtimeBatchrecover reliably.