I just noticed this on that wiki page:
" ...a worst case latency of up to 10ms (time between the moment an interrupt occurs in hardware, and the moment the corresponding interrupt-thread gets running)"
The useful number I need when I talk about the latency of a digital audio device (as a user of that device in a setting where such things matter) is the time between when the sound goes in to the device and when it comes back out.
That's not the same thing as the author of this wiki page is talking about. If the thread doesn't even wake up for 10ms it would have to do the processing and then deliver the output to the sound card for the next interrupt (or the one after that, or worse).