> 
> this is a well document problem on netbsd sparc.  [...]
Thanks for the info.

> i haven't yet tried out the new zs code gordon contributed.

I should point out, just so nobody gets any false hopes, that
the problems with excessive ZS interrupt latency are not the
fault of the ZS driver, and therefore will not be improved by
the new version of that driver. (no better, no worse)

Finding and eliminating latency problems is time consuming,
but that is the only thing that will help in this case.

Gordon