Thanks for the report. I've been building the server product on a stable code line branch, but one that is a few months behind the code line trunk. A couple of days ago, there was a point at which the frequently-unstable development trunk was stable (according to results of automated unit and system testing), so I created another branch from it. The new branch should have better handling of unavailable mounts (as long as the OS recognizes the mounts are unavailable). I expect to be releasing from that new branch pretty soon since it has been testing consistently well, although I'll first conduct a week-long stress test to see how the server built from the new branch performs under heavy load. What were the memory usages you were seeing with other products (if any) under the same conditions?