What — extend remove()'s not-found no-op treatment to purgeUnreferenced()'s own bucket.delete() call.
Why — a consumer running both the sweep and an independent retention job can race: the other job deletes a blob between the sweep's candidate scan and its delete pass. Today that surfaces as a recurring deleteFailed/error-log entry on every such race — exactly the noise class remove()'s no-op path (same file) was built to prevent. Fails closed (nothing wrongly deleted); noise-only.
Scope — modules/uploads/repositories/uploads.repository.js purgeUnreferenced(): catch the GridFS not-found error on bucket.delete() → debug log + count as already-gone, continue the pass; genuine bucket failures still surface. One unit test: mid-sweep concurrent deletion → no error log, sweep completes, counts accurate.
Surfaced by post-merge review of #4017 (2026-08-05).
What — extend
remove()'s not-found no-op treatment topurgeUnreferenced()'s ownbucket.delete()call.Why — a consumer running both the sweep and an independent retention job can race: the other job deletes a blob between the sweep's candidate scan and its delete pass. Today that surfaces as a recurring
deleteFailed/error-log entry on every such race — exactly the noise classremove()'s no-op path (same file) was built to prevent. Fails closed (nothing wrongly deleted); noise-only.Scope —
modules/uploads/repositories/uploads.repository.jspurgeUnreferenced(): catch the GridFS not-found error onbucket.delete()→ debug log + count as already-gone, continue the pass; genuine bucket failures still surface. One unit test: mid-sweep concurrent deletion → no error log, sweep completes, counts accurate.Surfaced by post-merge review of #4017 (2026-08-05).