I am operating my device in LMFS=82820 mode and JESD data arrives exactly as seen in Table 15 of the spec. Does properly functioning deterministic latency guarantee that samples from each of the four sub-ADCs (I don't know the correct terminology here) always align the same way within the JESD frames after each power-up?
On my hardware I believe that this is not the case because I have a spur correction algorithm that requires me to shift coefficients (or samples) after each power cycle. The coefficients are initially created assuming the sub-ADC alignment is always the same.