Rick,

Here's the results, somehow I was off by a factor of 10 though, VFP only got through 60,000,000 objects during the test. Each pass does 24 objects constructs/destructs. The results are from a query using group by iMemory

iPassCount tCreated              iMemory
       0   11/01/98 06:49:20 PM  281556
       1   11/01/98 06:49:21 PM  299076
       2   11/01/98 06:49:21 PM  299116
       3   11/01/98 06:49:21 PM  299108
    1910   11/01/98 06:50:07 PM  299144
  206223   11/01/98 08:13:25 PM  301708
  452947   11/01/98 09:54:00 PM  304264
  699671   11/01/98 11:35:39 PM  306824
  946394   11/02/98 01:17:51 AM  309392
 1193118   11/02/98 03:00:03 AM  311944
 1439842   11/02/98 04:42:49 AM  314504
 1686566   11/02/98 06:26:38 AM  317068
 1933289   11/02/98 08:10:25 AM  319624
 2180013   11/02/98 09:55:35 AM  322184
 2426737   11/02/98 11:40:36 AM  324744
 2465288   11/02/98 11:57:08 AM  324712

So sys(1016) does report a very gradual, the 2k jumps at about every 5,000,000 objects, increase in memory use. But that might happen normally while VFP is executing for that long a time. I'd have to run another test without objects to verify that. I wouldn't really classify this as a memory leak per se though. And this is completely artificial because I doubt very many VFP apps go through that many object instantiations during their run time.