...well. I did another little test. Saved 3 versions of file with 5 figures, trimmed down to 30 frames. One version with all the keys generated by walk designer, another version with 90 percent of the keys edited out and a third version with no keys. I did this to make sure I wasn't fooling myself with expectations from the first tests, not to mention possibly bad observations from memory issues that could have been created by using Poser or other apps first and not rebooting in one test or the other previously. Anyway, on my system there appears to be no noticable difference between them for scrubbing purposes. It's still good policy to knock out unneeded keys, of course, but doesn't look like it's worth doing if your only aim is to improve scrubbing. I'd like to see the same tests done with a higher end graphics card though.