>>Hi Bruce
>>OK, if you ever do find some resolution or info about leaky scans post something on the UT; because sometimes I get into huge loops in statistical apps ( sometimes I work on cross correlation of results of 100,000 students with 10 exam results over 250 subjects :-)) - can take hours.... as you know.
>>Regards
>
>I tested scan and count in simple situations over large numbers of loops while watching football yesterday. The longest runtime, though, was only about 1/2 hour, not as long as the time for the leak I've had show up, though I actually ran more total loops yesterday. But two things were noteworthy:
>
>1) There was no slowdown for either one. In fact, the opposite was true, they both took longest per/loop times when starting, then got faster and faster over time (weird?) until they eventually flattened out after some minutes.
>
>2) Count was *much* faster than scan for all tests, a good 30+% faster. The moral is always use count when possible...
>
>So there is some other factor causing my problem, and it may or may not be related to scan (maybe scan calling a complex method repeatedly). But for now, a simple scan appears fine.

I appreciate that info,
Thanks - muchos gratias!