Hacking up the BRM structs to make them use IDBDataFile consistently, I noticed that the EM, VBBM, and VSS are still using my v 0.1 prototype code for loading and saving. That code, written like 12 years ago, loads and saves elements one by one. That won't scale, and it'll cause noticeable pauses across the system while it does that. This is another one of those things I had to fix at Tune (saving a 600MB extent map ~50 bytes at a time = pain).

This task is for making them save consecutive groups of elements at a time. Much much quicker.

One other thing I noticed is that it's taking a full snapshot of all of the structs on commit. That's expensive, esp with the current impl, and it doesn't need to do that.