fixing bug 469448 (or, rather, fixing it w/o breaking tests) led to a regression where we opened up folder dbs for sub-folders on startup. This fixes that regression.

the fix is to move the code that forces us to open the db into the same if clause that we hit if the folder doesn't exist (which is what the test code runs into).

fix checked in.

likely Bug 491130 is fixed by this. (can't remember if bienvenu and I discussed tests on IRC, so....) Is the best approach to detect regression like to rely on TS tests? If so, we need to make sure TS tests have sufficient subfolders. And if not, do what type of tests do we need?

Do the TS tests actually create msf files for folders, or run against profiles that have .msf files? This bug in particular was about opening .msf files that existed on startup, which is something that afaik, none of our tests are good at catching.