Asset pipeline can't record its own error: foreign key violation on deleted/missing project

Error

*errors.withStack: could not save pipeline error to asset: ERROR: insert or update on table "assets" violates foreign key constraint "fk_projects_assets" (SQLSTATE 23503)

Stack trace

Alert (monitoring/alert.go:29)
  ← (*DaemonRunner).collectErrors.func1 (daemons/daemon_asset_pipeline.go:136)

Underlying cause chain: *pgconn.PgError (SQLSTATE 23503) → could not save pipeline error to asset: ... (same message, wrapped)

Request context

• The asset pipeline daemon (collectErrors) tried to write a pipeline error onto an assets row, but the update was rejected because it violates fk_projects_assets — i.e. the asset's referenced project_id doesn't exist in the projects table (likely the parent project was deleted concurrently, or the asset row references a stale/nonexistent project).
• This is doubly bad: not only did the underlying pipeline step fail, but the failure-recording path itself throws, so the original error content is lost and only this secondary FK violation is surfacing in GlitchTip.
• Worth checking whether asset/project deletion should cascade-delete or block deletion while a pipeline run references the asset, and whether collectErrors should tolerate a missing project gracefully instead of crashing.

GlitchTip

• Issue: https://error-tracking.devguard.org/devguard/issues/3615
• Event ID: 019fc51a3595754e8c54d78fd5170fbd
• First seen: 2026-08-03T00:50:50.069Z

Start in monitoring/alert.go:29 and daemons/daemon_asset_pipeline.go:136, then inspect DaemonRunner.collectErrors and the fk_projects_assets relationship. Reproduce or trace the missing-project path and determine how deletion and error recording currently interact. Done means the original pipeline error remains available without the recording path failing on a missing project.