>>> [...success with IPCs...]
>>> Are there hardware/software factors that might be affecting the
>>> stability (other than I don't push them to the limit)?
>> Based on my own experiences, I suspect [paging activity].
> Someone had a 'sledgehammer' patch which added explicit cache flushes
> to various places which seemed to almost eliminate the effect (at
> some significant performance reduction).  Would it be possible to
> test adding such flushes to various points in the paging code?

I don't see why not.  I'd certainly be happy to do what I could to help
someone doing such work.

Or do you mean would it be possible for me to test it?  I could dig out
my most reliable test cases and a crash-and-burn box.  I doubt I can
assemble the round tuits (time &c) to actually do the work myself, but
I could set the box up and give (eg) you ssh serial console access;
there might be problems, but, I'm confident, nothing insoluble.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B