Run sync jobs in one run, under one corpus lock, concurrently where that is safe
(planLanes on the jobs' parliaments). Every job's window is checked once, before
the lock is taken or anything is asked, and no label may come twice
(jobListProblem).
One job failing does not stop the others, unless stopOnFailure: its error is its
outcome (status: "failed"), and the caller decides what the run as a whole exits
with. An aborted signal stops each running job between two refs, as in sync(),
and the jobs whose turn had not come are skipped.
Run sync jobs in one run, under one corpus lock, concurrently where that is safe (
planLaneson the jobs' parliaments). Every job's window is checked once, before the lock is taken or anything is asked, and no label may come twice (jobListProblem).One job failing does not stop the others, unless
stopOnFailure: its error is its outcome (status: "failed"), and the caller decides what the run as a whole exits with. An abortedsignalstops each running job between two refs, as insync(), and the jobs whose turn had not come areskipped.