Do the trace, but also make sure your don't have XON/XOFF, as that can break some messages because it changes bytes. Because the code in the IAP passes the Modbus through without analysis, it's impossible it cares differently about 40001 or 40616.
As userid suggests, you need to do the ia trace because 1 of 3 things is happening here:
1) the IAP is sending a request & the slave isn't answering
2) the IAP sends request, slave answers, and IAP drops response as an error.
3) the IAP sends request, but slave answers too slowly and IAP has already given up.
For #3, some slave devices use multiple PCB talking internally. For example a Flow Computer might have a main comm card, which answers "its" data in 15msec, but if you ask for a totalizer value, it uses a backplane to ask a second card, so the answer might take 780msec. So different registers cause different timing behavior. If this is true, you MIGHT see in the IA trace some form of 'unexpected' or mismatch error show.