Here's how it was described years ago by someone from SCS (and now reinterpreted by me from memory). Vegas caches rendered frames in preview RAM when it can't simply play the footage outright. Then on the second pass through the footage it only needs to render frames that it missed on the previous pass. In this way each pass through a region of the timeline gets a little better. If you were to throw enough processing power at Vegas then it ought to get all the frames rendered on the first play.

Looked at another way, every frame you ever see in Vegas must have been read from disk and decompressed (rendered) to RAM. If Vegas can't keep up then it has to skip frames, and if it skips frames then it doesn't purge the previous frames from RAM but instead saves them, presumably until there's no room left in your preview RAM space. This improves playback of the region because fewer frames need to be processed in the next pass. How well this works with a long GOP might be an interesting question, the behavior dates back to when DV25 was Vegas' bread and butter.

I've often wondered if Vegas' caching frames from timeline positions into RAM had something to do with random frames and black frames appearing in renders, but if that's a problem it's probably not the only problem.

Rob