LACP LAG GSM7248V2 with another switch not aggregated

So I was lucky to have a bit of a progress here. As the situation allowed me to reboot both switches, I did so and the LACP started to work as anticipated. Both interfaces were added to the trunk on both sides, everything seemed to work properly...

...well, except for NFS. I have an nfs server connected to one of these switches and several clients connected to each of these two switches. Those clients that are connected to the same switch as the nfs server (thus, not relying on the trunk between the switches) work flawlessly. However, those clients that are connected to the server only through the switch-switch trunk, show all kinds of weird behavior. Mismatch between the amount of data sent by one party and received by the other. Lagging data transfers, "server not responding" type of messages, etc.

All the other protocols I tested seem to work just fine. DHCP, IPMI, ssh, ICMP, TFTP. Packets do not get lost, the performance seems to be stable. Just the NFS does not work.

This suggests that the individual lines are working fine and once the connection is established and associated with particular cable, everything is fine. However, NFS requires multiple simultaneous connections for all the related RPC services. It seems that there is some issue with their synchronization when different physical lines are used.

I tried both, static and dynamic trunking, with the same result. BTW., the other switch involved in this trunk is a single SMC TigerStack II 8848M.

I would be very grateful for any hint.