DavidR,
Thanks for that. I did indeed read both articles multiple times before posting. They seem to be the most relevant for my query, but I still found myself unsure about my understanding of them. Here was my reason for posting the questions (enumerated based on the same question numbers as my original post).
1) It seemed like the answer is yes, but being unfamiliar with what small set of files go into the cache (and with Windows under the hood in general), I was hoping to confirm my understanding. In particular, I was not familiar with the factors that cause such files to be so resistant to malicious modification or replacement as time wore on. I'd still be interested in a layman's summary of this, if such an explanation is even possible.
2) Same reason as for (1)
3) My take on this article (from a layman's perspective) is that it explains speedups from one implementation of persistent cache to the next, but not the speedups from using persistent cache versus not using it. Empirically, the article implies there is speedup, but I was wondering more about the reason. With very few files qualifying for caching, it would seem that they must be very big in order to impact the scan. This is "obvious" from a nonexpert viewpoint, and I was seeking to sanity check it.
4) I'm was wondering why the default was to spend time populating the cache but not using it.