We just got the following report in from a server (9.0.1 FP 3) with 2000 users on it. Lots of data, and long backup times.
It's a mess, with nearly 14 million fragments (one NSF file has over 173 thousand).
The good news is, here we have an opportunity to improve performance in a huge way!
They didn't know... and now they do.. check to see what DFA might be able to tell you!