[09:13] <cjwatson> wgrant: Job leases confuse me.  Do you know of a particular reason why queueing a job shouldn't just release the lease?  AFAICT retry_delay should be sufficient to arrange for it to be queued a reasonable time in the future, and having the lease still held just confuses things.
[09:14] <wgrant> cjwatson: retry_delay is only for celery, isn't it?
[09:14] <wgrant> I agree it's confusing, but it's not *just* confusing.
[09:14] <cjwatson> Oh, does nothing else honour scheduled_start?  Hmm.
[09:15] <wgrant> Oh, it's possible
[09:15] <cjwatson> Job.ready_job checks scheduled_start
[09:15] <cjwatson> ready_jobs
[09:15] <cjwatson> So surely a job won't show up in iterReady() until the retry delay has elapsed.
[09:16] <wgrant> You are probably not wrong.
[09:16] <wgrant> It was all slightly strange, and I ripped out some strangeness a few years ago, but not all of it.
[09:16] <cjwatson> (I'm trying to fix the weird snap store upload retry behaviour.  I have it mostly fixed, but it's a bit of a hairball.)