Yes, I don’t think that we got to wait latency like on DT2, or DN2.
I’ve played a lot with RT kernel + jackd(as more stable audio framework) on RPi5(ARM platform too) for different purposes, with own builded RT kernel and optimised Linux env, but I never got audio loop better than 8ms(with overwitch or or some jackd clients like usb audio emulated device to connect to iPad as host), in fact 8ms was a best result without xruns.
But it’s acceptable for me.