Slave not working when master down

[lists at rak.radio.cz]

Hi,

on our network, I have two DNS servers plus one outside of our network. 
Local DNS servers are a master and a slave and the remote one is slave as 
well.

The problem is that when the master goes down for whatever reason, almost 
ALL traffic on the network stops. We have linux workstations and at that 
moment they freeze altogether. Could the problem be NFS, because they all 
use it?

Anyway the question is - shouldn't the clients use secondary slave 
for nslookups when the master is down? The clients are properly configured 
- they have both DNS servers listed in /ect/resolv.conf - configured via 
DHCP. Still it seems that when the master doesn't work, workstations don't 
use the slave. And what's even worse - our network gets unreachable from 
the Internet. Root NS records are proper - they point to the master and 
the outside slave.

Both slave and master are running BIND, master is 9.2.0 and slave 8.2.2 (I 
know, I am just about to upgrade...).

When I query the slave manually, all works fine.

Any idea or hint will be greatly appreciated!

Jan