TL;DR: if jack2 audio becomes corrupt after upgrade, then increasing the jackd buffer size or periods to fix the audio doesn't necessarily increase the observed latency from before the upgrade.
I've been using jack2 with low latency settings on a Scarlett USB audio i/f:
$ jackd -d alsa -d hw:USB -r 48000 -p 128 -n 2
I want to keep the latency low for live performances. That was all working fine until kernel 5.15.7, which includes some latency improvements to the "ALSA: usb-audio" subsystem. The audio output became corrupt and almost inaudible after the upgrade. I thought it was a bug, and after some consultation, was given a workaround that disables the improvements using boot option snd_usb_audio.lowlatency=0.
But... it turns out that my original settings only worked before because of an induced hidden latency in the USB audio subsystem that was fixed with the kernel changes. Basically, my system couldn't cope with those settings (buffer size and periods) once that hidden latency was removed. I had to bump the periods to 3 to get my audio back.
In fact, the real world latency, as measured by jack_delay, went down from 14.2 ms (old kernel) to 12.5 ms after increasing the periods to 3 with the new kernel:
$ jackd -d alsa -d hw:USB -r 48000 -p 128 -n 3
So all is good.
]]>