We have a NFS3 share on a file server mounted as homes on our clients and we sometimes (depending on use, between several times an hour and once a week) experience short freezes that we had a really hard time pinpointing.
After some long and painful debugging i think i found the solution in this notice from our cisco firewall that corresponded directly to a freeze:
Feb 28 11:18:34 192.168.98.233 fw56 %FWSM-4-209003: Fragment database limit of 200 exceeded: src = 10.153.59.2, dest = 10.153.59.1, proto = udp, id = 57950
It turns out this is a known problem that I was unable to google: http://www.cisco.com/en/US/docs/security/pix/pix63/system/message/pixemsgs.html#wp1053420
Hope this helps someone, it took us months to find it because we had so many valid ideas for the possible cause. I now setup some work stations to use nfs4 and some to use nfs3 over tcp. Let's see what works better.