David,
Now,,,,,I understand somewhat, generally, what the problem is. Is there someway I can change the limiting value, and possibly the specific memory size, without a smoke and burn?
No, but I can create an experimental version that will allow an integer math relolution of 3000x3000 and email the executable to you. My experiments with this version did not reveal any obvious problems. I will need to look closely at the code before I consider making this a permanent change. It might make sense to tie the switch to floating point to magnification as well as resolution, or some combination. Let me know if you have a high speed connection and/or don't mind getting an email with a 588 KByte attachment.
btw: What did you think of the pix? Or could you tell the problem just from looking at the pars?
I liked them. Especially sf700233. I made a par of one of the "bad" images, then compared it with the original. The new par had float=yes and the original didn't. Jonathan