>... just leaving stathz=0
>Would anything important be lost with this approach?

Oh, only accurate and reliable statistics and fairness. :-)

(See the Usenix paper Steve McCanne and I wrote back when we
did the statclock code.  It contains a `hog' program that will
suck down over 90% of the cpu time on a SPARC box running SunOS,
yet shows up as using less than 2% of the time, and it makes
the machine virtually unusable.  The same program is spied
out by the separate statistics clock method.)

Chris