> Could some of this be an effect of caching? Running the first query
> will get all the interesting parts of both tables into the cache,

making

> the second one quick.

Good theory Ann. That is my money now. Disconnecting and reconnecting
between the two queries should effectively clear the cache. I
personally don't find execution time a reliable measure of query
performance because of caching. I just make sure there are no
unneccessary table hits (or as few as possible).

Adam