BUG: config maxConcurrentCallsPerWorker:1 drops another call if worker crashes

[Akuukis]

Problem

My use-case is both cpu and memory intensive calculations with expensive initialization, where some of them are too big and throws heap out of memory, or too expensive and hits a timeout, which are pointless to repeat and therefore I use maxRetries:0. I want to be able to re-initialize only crashed workers, and make sure that those crashes doesn't skip or abort other calls.

This hint in README was misleading - it suggested that maxConcurrentCallsPerWorker:1 would save my day, but it didn't. Instead, per each crash 1 another call was skipped/aborted, or all of other calls when used with 1 worker.

[..] Preferably run this with a maxConcurrentCallsPerWorker so you don't interrupt other calls when you have a timeout [..]

Workaround

Use maxCallsPerWorker:1 and maxConcurrentCallsPerWorker : 1. It re-initializes every worker every time, but at least doesn't skip/abort other calls.

Expected solution

I wish maxConcurrentCallsPerWorker:1 would work with my problem as I imagined.

Comments

See linked PR for working tests that use the described workaround.