sdexec: transient units may not be getting cleaned up properly

[garlick]

I've noticed a few transient units are not getting cleaned up. For example:

[root@tioga20:garlick]# ~garlick/bin/flist
  UNIT                          LOAD   ACTIVE SUB     DESCRIPTION              
  dbus.service                  loaded active running D-Bus User Message Bus   
  imp-shell-9-f2yHNSwmAhHR.service loaded active exited  User workload         >
● imp-shell-9-f2zEnyt2ai3q.service loaded failed failed  User workload      

The first job died with a "failed to create guest ns" exception

[root@tioga20:garlick]# flux job eventlog -H f2yHNSwmAhHR
[Apr12 10:55] submit userid=27241 urgency=16 flags=0 version=1
[  +0.018449] jobspec-update attributes.system.bank="guests"
[  +0.018531] validate
[  +0.037056] depend
[  +0.037109] priority priority=623
[  +0.123069] alloc
[  +0.123345] prolog-start description="job-manager.prolog"
[  +0.123398] prolog-start description="cray-pals-port-distributor"
[  +0.282366] prolog-finish description="cray-pals-port-distributor" status=0
[  +1.222043] prolog-finish description="job-manager.prolog" status=0
[  +1.232569] start
[  +1.605169] memo uri="ssh://tioga20/var/tmp/lbannusr/flux-7chyxa/local-0"
[Apr12 12:26] exception type="exec" severity=0 userid=763 note="failed to create guest ns: No such file or directory"
[  +0.000119] release ranks="all" final=true
[  +0.000150] free
[  +0.000172] clean

The second job is not a valid job id.

I did do a sanity check and found no units left behind when

• a job succeeds
• a job fails
• a job is canceled