Looks like the place for the following hard(?) to spot issue...
If you try using one of these music packs with a user map other than E1L8.MAP, you'd hear that it's not working and plain MIDI is played back.
Apparently, if you load, say, a map named PLATINUM.MAP, then it'd look for PLATINUM.MID or PLATINUM.OGG first (probably the OGG beforehand), and if both are not found, it'd revert back to the "vanilla" E1L8 track, MIDI version, skipping the OGG one.
Some more relevant info: Say that you load DN3D as-is with no arguments (with the possible exception of the music pack), and then select E1L8 from the "User Maps" submenu. Then you'd hear the OGG track as expected. Now select some other map and you'd hear the MIDI. Select E1L8 again, and you'd still hear the MIDI.
On the other hand: Load a non-E1L8 user map from the beginning via command line arguments, and you'd hear the MIDI; Then select E1L8 from the menu, and you'd hear the MIDI again.
Now when I think about it... E1L9 to E1L11 may also be affected in similar manners.
Regarding these, there are two user maps, both named E1L9.MAP (yep...). Vanilla Duke behavior would be to use the E1L8 track (dethtoll) if these are loaded with "-map E1L9", and the E1L9 track (streets) if "/l9 /v1" is used ("-l9 -v1" for EDuke32).
Truly, interesting... guess there are wrong logics which can be fixed, although it's maybe more complicated than what one may think (especially for maps like the two "E1L9"s).
Well, that's it for now