crashing with running jobs can lead to problems

[garlick]

Problem: if the flux rank 0 broker crashes with running jobs and then restarts, the fluxion scheduler may fail when processing the hello protocol.

2023-01-17T16:10:17.222950Z sched-fluxion-qmanager.err[0]: hello: error loading R for id=202046916802905088: No such file or directory

To get fluxion loaded, those jobs have to be removed from the KVS.

Until we can handle recovering running jobs, we should probably just force these jobs into inactive state.

Caveat: tasks belonging to the job could still be running.

Edit: the fluxion bug that is triggered here is flux-framework/flux-sched#991

[garlick]

I think #4894 addresses the short term issue here so closing.