Thanks a lot for all replies. I have increased the number of nmbclusters 
in loader.rc, as follows:

1. on generic-PC to 64 MB which has 256 MB RAM:
set kern.ipc.nmbclusters="32768"

2. on my WRAPs to 32 MB which have 128 MB RAM:
set kern.ipc.nmbclusters="16384"

Since the error had not reocurred, it looks as if they have plenty of 
headroom now. Fortunately, because the systems were hanging and required 
a manual reboot.

Not sure if the Watch Dog on WRAP would have caught that problem, but I 
am about to get the SC1100's Watch Dog code in the kernel to work anyway :-)

Rolf