Found the memory corruptor. String trimmer didn't have a valid

termination test.  Sheesh.  Also get some more numbers out of the
example/load test program which drives traffic.  This should
give some useful insights into how the current http throttle
works (or doesn't) with varying client demands.