Jira Align Connector and Jira Connection Pool Exhaustion: Atomic Rollback or Partial Commit During Sync Interruption?

The Jira Align Connector exchanges work items with Jira over REST, yet the transaction-boundary behavior during Jira-side connection pool exhaustion is not clearly specified in public materials. When pool acquisition fails mid-batch, it is uncertain whether the connector rolls back atomically or leaves partial Jira issue updates that could produce duplicate links or status mismatches.

Because Jira Cloud surfaces API throttling as rate-limit responses while Jira Data Center exposes local pool metrics, identical symptoms may require different diagnostic starting points. Connector error handling also varies by release, so behavior cannot be generalized across versions without checking release notes.

Is the rollback-versus-partial-commit behavior for interrupted sync transactions documented, or is it undefined and environment-dependent? Does the installed connector version dictate whether failed batches retry, queue, or partially commit? What evidence distinguishes an environment tuning problem from a connector defect when pool exhaustion interrupts a sync batch?