I've heard rumors that we should consolidate our DGNLIBs into one?

Most definitely.
The truth is, Bentley can suggest whatever they think is best, but I will nearly always default to what works best for my organization. I've been ignoring them (about some things) for years...

this is an issue for those developing and maintaining workspaces for the client.

Bentley did recommend that we consolidate our individual drawing seed DGNLIBs (we may have had 20 or so) into a single DGNLIB to speed up the workspace load time when opening ORD. Which we did. You can see that it's not loading all 20 now, instead only the one. So, some performance increase was noticed. I did see or hear this was being promoted for drawing seed DGNLIBs in different communications, coffee corners, etc. (I don't recall exactly which) so maybe this is what you've heard. 

Even if I hear something from Bentley, I'll continue to do what I'm doing - using my clients' standards as delivered.

With the recent recommendation by Bentley to merge all drawing dgnlibs really got me questioning their approach.  Also their recommendation of defining folder locations for faster download in PW.  I have been managing workspaces for 23 years now and simpler is always better.  You guys have me wondering about character length now too. Microstation J would crash if dgns or resources had too long of character path lengths and I want to say its limit was 176 characters.  There was a reason to our madness of short folder names and file names.  :-)

A great test to try is run ORD without a workspace on Projectwise.  It screams!  Logically the slowness would be the workspace resource files then. So logically removing as many resources files as possible would increase the load speed and processing speed. Right?