Another Mastering/ASIO observation and question

During the mastering process I describe in the post below, we used the ASIO loop (AES digital I/O via Lynx AES16 card) a number of times to process files through a very expensive, very high quality digital box. The other engineer involved is a well known veteran with a well tweaked computer.

We had done this before on an earlier project, too. In both instances, while the majority of the resultant files were fine, I found glitches in them that had to be dealt with. In the final case, I ended up having to abandon the loop and feed to an external capture machine instead.

Considering that the box we were using has an inherent 96 millisecond delay, and considering that there was a fair amount of AES cable going to and fro the machine, I am having trouble understanding how the ASIO loop could really be expected to keep the incoming signal in sync with the outgoing, which is operating on the central clock for the machine.

Raising the buffers to the maximum did not fix the problem.

Does anyone have any insight into ways of using the ASIO loop that can avoid these glitches in the future? I probably will avoid it from now on anyway, but would like to have a better technical understanding of how the loop is supposed to maintain sync in real time.

I’m not sure the External FX loop is there to do “realtime in sync” editting

for me, it’s going OTB for “chain processing” and back ITB for printing
I have to set my RME ASIO to 512k in order to do this without potential glitches. 256 works most of the time, but 128 is distorting.