I'm seeing a production server which is peaking at over 100 threads and with occasional queue length spikes of 20-30. The underlying hardware has plenty of excess capacity.
I'm wondering if the self-tuning thread pool has a default max number of threads so it isn't increasing beyond this limit.