Resolved. DNS lookups from production were slow between 18:10 and 18:32 UTC, and requests that needed those lookups timed out.

Workflow runs and batches were delayed, and some failed to save. Data syncs to connected systems failed during the same window. Most of those writes retried successfully once DNS recovered. A smaller set that used up its retries may still be missing; we're checking those.

No action needed on your side unless a run or sync from that window looks incomplete.

Sorry for the interruption.