Files
lefarcen bfa305f17a fix(web): resilient retry for workspace-context writes (#6604)
Extract a shared BackoffController primitive and use it so
workspace-context-dependent writes survive transient outages instead of
failing closed.

- resolvedWorkspaceContextForWrite proceeds on a generation-matched last-good
  context instead of throwing on a transient `unavailable` (a weak-network blip
  no longer bricks project creation and every other workspace write); a cache
  from a retired generation (account switch) still fails closed.
- context + billing refetch use exponential backoff + jitter, single-timer
  coordination, success-only reset, and SSE early-trigger without depth reset
  (replacing billing's fixed 5s and context's no-retry).
- createProject auto-retries a 503 retryable WORKSPACE_AUTHORITY_UNAVAILABLE
  with the id minted once (idempotent).
- unify the 3 duplicated SSE reconnect loops (project-events, Theater,
  useEventStream) onto BackoffController and add jitter.
2026-08-07 10:55:18 +00:00
..