When all available threads are busy, test job will be queued for 10 minutes (test client will wait for HTTP response for 10 minutes or will fail with socket timeout exception if timeout is less than 10 minutes). For us this functionality causes problems: users are waiting for 10 minutes and more to receive the concurrency error.
In our cases limits may be set on the parent account level. So multiple accounts will share the limited number of threads. In this case it's not possible to control all accounts and be sure that nobody else uses SL threads. It means that we need a way to "fail fast".
Proposed enhancements for queueing mechanism:
- implement ability to manage queue wait timeout
- add possibility to disable queueing mechanism (maybe just setting wait timeout to 0?)