Can a failed Bamboo deployment be retried on the same release without duplicating external writes?

In Bamboo, a deployment project consumes a release created from a successful build result. When a deployment fails after a script or deployment task has already written to an external system, the operator must choose between redeploying the same release and creating a new release from the same build result.

Bamboo records deployment outcomes per release and environment, but it does not provide a general idempotency token or transactional rollback for arbitrary tasks. Retry safety therefore depends on whether the target operation is idempotent. Manual stages and final tasks can gate promotion or run cleanup, yet they do not compensate for partial external effects. Release naming and downstream triggers may also change if a new release is produced, and behavior around rerun options and release immutability varies by Bamboo version.

Does redeploying the same release re-execute all deployment tasks and therefore duplicate external writes unless the task itself is idempotent? Which Bamboo configuration options—such as rerun behavior, manual stages, or final tasks—help prevent duplicate side effects during a retry? When should operators create a new release instead of redeploying, and what impact does that have on release identity and downstream deployment triggers?