Retry storms that paint green tiles

A silent backoff loop can look like healthy sync when you only read success counts.

Close view of illuminated data centre hardware

On sync endpoints, a client that retries without jitter can multiply catch-up work for a single user action. The success counter climbs; the product never asked those users to wait—or to leave.

Before rewriting the feature, walk the path while watching live sync and error rates. Note client-only retries versus server confirmations. Check whether staging shares production reporting keys.

Conflict & Drift Reporting stays narrow on purpose: one reporting cycle, deep reconciliation, a checklist you can rerun after every release so the green tile does not return as fiction.

All field notes Talk through a similar issue