Comment 4 for bug 2062568

Ran into this again just hours after commenting by attempting to unpack a large archive file. Apparently the new setup of higher performance host with more network bandwidth is just too overwhelming to be usable with NFS with this issue.
Reading alone still seems to be fine though.

Torturing 50% of the memory with memtester didn't reveal any problems, and getting the server to do the unpacking of files does the same mixed I/O with no issues so far, progressing way beyond where I could get over NFS, so it doesn't look like an HDD issue either.

The info dumping described here seems to be potentially useful for the next catch, so linking it here:
https://www.spinics.net/lists/linux-nfs/msg97486.html

At some point I may try to reproduce the issue with just writing to see if the mixed workload is really required for the freeze, but not sure when will I get to it, can't just restart NFSd, and the mandatory reboot to get it going again is only feasible when all other tasks running on the host can be interrupted.