You'll probably have to just use a subset -- try it and see. We're planning on replacing the backend-database with something a little more scalable.

The (current) plan is to ship the next version sat on top of a Microsoft Access database, but since it's probably gonna use ADO, it'll run on SQL server if you want to hack with it a little. The server running on my machine runs on SQL server with approx 25k MP3s in it.

The player still suffers from instability with a dataset that large -- it's the first thing on the list to get fixed when we get around to doing the next release.