I had a similar problem not so long ago.
Then I just had not assigned enough memory to nss. Then I gave NSS 500 MB
and all worked fine for a while. But I watched my cluster (nw60 with sp5)
all the time and discoverd that, if someone stores a large file on the
server, there is not enough cache memory left. I used netware remote
manager to see what happend and took a screenshot of the Server Memory
Pools Summary Screen.

Here are the highlights:

Server Memory 3839 MB

Free Clean Virtual Memory 2370 MB
In Use 35 MB
Virtual Memory Total 2405 MB

File System Memory 947 MB
Reserved Swap Memory 38 MB
Swap File Size 40 MB

Code 22 MB
Data 13 MB
Alloc L!=P 343 MB
NLM Total Memory 380 MB

Cache Buffer Size 4 KB
Original Cache Buffers 3835 MB
Current Cache Buffers 102 MB

When cache memory goes below a certain value, network connections are will
be terminated and sometimes the server will abend.
As I am running a cluster on a shared fibre channel storage, I think it
could probably be a problem with my HBA's.
Why is there so much unused virtual memory?

Can anybody help me?
Greetings Stephan