health check reported hung sagas that aren't actually running

In our dogfood environment, when we first deployed oxidecomputer/console#3226, we noticed the "contact support" bit was set. The reason was a bunch of sagas that have been running for a long time. Investigation of one of these sagas revealed a case we didn't consider with the health checks: a saga that (1) has not finished (it's still unwinding), but (2) is not running and can never be running in Nexus again (because a steno bug is preventing it from being recovered).

Questions here include:

1. Should Nexus handoff during update block on such sagas (i.e., not proceed while there are sagas that couldn't be recovered)? I'm thinking "no" because these aren't running and won't be running on the other side of the update, either. But it's definitely a dicey situation.
2. If not, should the health check still identify these? I'm inclined to say "no" because although they are a problem, they're not related to update (again, assuming we decide handoff should continue ignoring these).
3. What do we do to repair a system in this situation? (This question is not update-related.)

Start by tracing Nexus health checks and update handoff behavior for unfinished sagas, then review how steno recovery determines whether a saga can run again. Clarify whether unrecoverable, non-running sagas should block handoff or appear in health checks, and define the repair path. Done means these questions have an agreed behavior and corresponding implementation scope.