YES. That is exactly what I mean. You'd find a bunch of files you wanted and hit download. You'd get maybe three of them. The rest would "need more sources". You'd hit find sources. It would find a few more (usually three) and download them. The rest would end up "awaiting sources". If you selected them and hit resume, you'd usually get some more -- again, frequently, three. (Three seems to be a common number of upload slots, along with five, for whatever reason.) Wash, rinse, repeat, eventually get them all, or most of them. Now, the later stages of this method for getting the rest of the files Doesn't Work. You have to keep restarting Limewire, which is pointless and which people shouldn't be being encouraged to do, whether explicitly or by the program's behavior.

It is clear from what you say that "awaiting sources" is never supposed to appear for a file with known sources online; it is equally clear from what I have observed that files with known sources do erroneously end up in that state, despite being resumable. Except that in most of the 4.9 betas it doesn't work anymore, and as a result when you find a big cluster of co-hosted files you can only ever get a handful of them without restarting Limewire, which didn't used to be the case if you knew what you were doing. (It worked for some of the betas, around 4.9.10 or maybe 4.9.7, though...)

The proof is in the pudding -- if you find a batch of such files now and get a few and the rest end up awaiting sources, and you restart your client and it promptly downloads some of them (or does so shortly after you do another "find sources"), this is a sure sign that the host with those files never actually went offline, and the supposedly "awaiting sources" files were actually still available and should have shown "waiting for busy hosts".

Let me guess -- your new udp thingie cannot distinguish between an offline host and a busy host, while the old system could? And moreover, once it decides a host is offline it actually forgets about the host entirely, whereas the old system still remembered the potential source and if it became available again then resume would work, but now it won't because even if the source is available again it forgets it ever existed? Considering how common it is for hosts to be intermittently busy, in yo-yo mode, or have flaky connections, this is a recipe for disaster.