I've always worked in a relatively small environment with respect to users and feature classes/tables, but have deployed SDE/Enterprise Databases since v 8.x. My users though have read-write access to some of the feature classes and are physically located through out the valley.
That said, I recently took a part time gig with the county, and they have a 'public' access SDE along the dimensions you describe, with maybe with a larger user base. They use feature data sets to organize the data. At first glance I was thinking, 'Hmm... Feature data sets... Warning Will Rogers...' I know feature data sets were not designed as a file management system, but for a read only database, it seems to work for them.
In my primary postition (the smaller environment noted at the beginning) I mangage the data in an Enterprise db as noted. However, for specific uses (it's a 9-1-1 dispatch center) I typically replicate oneway to a FGBD. Perhaps such an approach could be depolyed in your case where you have several FGDBs of different themes of data. Depending on the level of sophisitcation of your users, they could replicate one way to thier own FGDB those feature classes and tables they typically use. [ That would keep you out of the replica mamanagement business, and put it on them.]
If replication isn't an option, I've also created models, and subsequent python scripts that use a series of Delete Features followed by Appends to empty out exisiting local FGDB feature classes and then append back in 'fresh' data. These can be fired as batch jobs at night.
That should just about do it....