I had a mr commit in /etc hang. Investigation showed it was trying to stat a path under /media/, to a drive that had automount enabled, but which happened to have its USB port turned off to save power. In this situation, any access to the drive's mountpoint can take several minutes for the automount to time out before it fails.

It seems unncessary for mr run in /foo to cause mr to stat repos in /bar that it's not going to be working in. I don't think it's worth going to major lengths to avoid, eg if a repo has chain=true then it's reasonable for mr to want to read its mrconfig at startup along with all the rest. But this repo in /media did not have chain=true, it was just an unrelated repo to the one mr was going to operate in.

Tracked the stat to the use of safe_abs_path in loadconfig.