Browser extension: detect and subscribe to standard.site publications on ATProto substandard.blog

fix(panoptiblog): retry repository reads from a backoff queue #596

Merged permadeath.com
targeting main from claude/pano-crawl-queue
Merged opened by and assigned to permadeath.com
targeting main from claude/pano-crawl-queue

Transient repository reads now go back on a work queue with exponential, jittered backoff. A failed item releases its worker while it waits; the phase finishes, the queue sleeps until the last delayed item is ready, then retries. A one-minute PDS stall no longer outlives three immediate HTTP attempts and drops the repository.

Canaries are read first and get eight attempts over fifteen minutes; ordinary repositories get six over ten. Only an exhausted repository keeps its previous records and is deferred to the next run. Permanent 400/404 responses are still skipped without retry.

The canary timeout that refused the 2026-11-09 run would have been retried through the stall instead of failing the whole crawl. npx vitest run scripts/panoptiblog passes 176 tests, including new queue tests for backoff, exhaustion, permanent failures, canary budgets, and deadlines. prek passes.

Commits 0