I do a lot of sequential tracking, so out of curiosity I rendered a click trick to audio in Studio One (yellow), and routed the audio out to my StudioLive Series 3 32, then to a mix which was patched back to studio one on another USB channel, and recorded it with different audio block sizes. Dropout protection was set to low. The results were suprising: the lowest block size had the greatest (albeit very small ~1 ms) latency, and larger block size started to run AHEAD of the source material. For 256 samples it was dead on. Seems almost as if there is a computation or rounding error in the latency compensation ...
I also did the same with the metronome playing directly (violet tracks), same results.
This is clearly inconsequential because it's so small, still, can anyone explain this?