[BUG] RR is losing jobs under load

[rauanmayemir]

I have a simple flow where I do stuff in a DB transaction scope, after which I send a job to the AMQP queue and finish the request.

I do understand that jobs are not guaranteed to reach the queue if server shuts down or client disconnects and cancels the request abruptly, but I didn't expect this to happen so often under a seemingly not very heavy load (barely a request per second on average, where request is a fairly heavy write op though).

I can expect to have about 10-20 jobs per day to be lost and not persisted into the queue. The worst part is that I can't track them down by noticing an exception or something.

[rustatian]

Transferred to the spiral/jobs repo.