Do failed jobs get cleaned up when jobs on the queue succeed? Or only when there is a new failure?

[simon-abbott]

I am confused about the minutia of how removeOnComplete and removeOnFail work. I understand that they are lazy, but I am unsure if that laziness applies to the queue (which is what I would expect) or the queue + status. From reading through the code it appears to be the latter, which seems odd to me, but I could very well be missing something—this codebase is complicated enough that I would not be surprised if I didn't trace it through properly. How does it actually work under the hood?

To clarify my question, say we have a queue with removeOnFail: { age: 60 * 60 * 24 } (expire after a day). If that queue runs one job every minute, and once a week that job fails, will that failure be kept for 24 hours (cleaned up when the next successful job finishes) or will it be kept for a week (when the next failure occurs)? If it is the latter, is that intended behavior?

[manast]

The removeOnFail policy will be applied when a new jobs fails (same for complete).