The logic of downloading was changed so that LimeWire would remember fewer sources between sessions, but those sources would be of higher quality.

This is a necessary step because now LimeWire quickly discovers pretty much all the hosts on the network that have had at least a part of the file at one time or another, and that number is usually very large. It would take too much room for LimeWire to remember every single download source it ever heard of between restarts, so it only remembers those it successfully downloaded something from.

It is not so much a sacrifice as a different tuning - LimeWire may seem to perform a little worse for downloads which are resumed after a restart, but will perform better for newly started downloads.