It seems that both use cases are the same: the agent or flow run
infrastructure crashes and the flow run and possibly also task runs in progress should be cleaned up (e.g. by setting their state to Crashed/Failed) to give more precise information that the current flow/task runs are in fact no longer Running but Crashed
I believe this existing issue addresses that
https://github.com/PrefectHQ/prefect/issues/5995, thanks again!