xRangerDEandClaude Opus 4.7
020f3dacf1
Two server-side changes touching different paths.
1. fetchPublicTwitchGql now retries on transient HTTP (408/429/5xx) and
network-layer failures (no response). Up to 3 attempts with
exponential backoff + jitter (400ms * 2^(n-1)). The previous
catch (e) { return null; } swallowed network blips on the public
fallback path, which is what every user without a client_id hits
on each VOD list load — a single TCP RST produced an empty list
and the user had to click refresh. GraphQL errors[] are still
returned without retry (application-level query rejections).
Recovery is logged via appendDebugLog so we can later see whether
the retries actually pay off in production.
2. shutdownCleanup() consolidates window-all-closed and before-quit.
The two handlers ran nearly identical cleanup blocks but had
drifted: only window-all-closed killed children and was
platform-aware. The helper kills activeDownloads + activeClipProcesses
+ currentEditorProcess with try/catch, persists config + queue,
then stops timers (debug-log flush moved AFTER persistence so any
save error reaches the log before the timer is gone). An idempotent
shutdownCleanupDone flag makes a follow-on event a no-op.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>