Square's 429 lasted seven minutes and 47 evening orders went to dead-letter — does your POS retry cron read the Retry-After header?

Mathilde (42), ops manager for a 5-location sushi chain in Lyon, opened Slack at 09:14 Saturday morning to "Square sync paused: 47 orders status='dead'." Forensic: 22:31-22:38 Square hit her chain with a 7-minute 429 (Retry-After: 60); the old retry cron didn't read the header, applied flat exponential backoff, and burned through all 5 attempts in 2.5 hours. PR #598 batch AAA F1: (1) Retry-After parse + MAX(exponential, hint × 1.2) safety margin + 1h cap, (2) per-provider intra-tick pause map releases remaining rows to `pending` without consuming attempts.

Related articles