It's been a while since I posted it. Basically, there's no universal
solution. There're several issues here.
1) Mount options.
It better to use bg and intr. Some1 suggested soft, but it's not good for
the following reason: suppose you run a program, which write to
NFS-mounted system. If you gracefully reboot your NFS server, the
program will wait 'til the client regains the mount and then will resume,
which is a nice behavior if yoo have jobs which run for days and weeks.
With soft options, you lose this feature.
Also, you should uncomment the offending line in /etc/mnttab.
Bottom line, it all depends on the nature of server downtime. My problem
occured when there were some huge hardware problems with NFS server.
Usually, mounts are regained if you have mount options as above.
Thanks to all who participated in discussion. I don't think this is a
final word on this subject, and if you have smth to add, I'd be very
grateful.
Igor