I'm still hoping there's a way to do this without creating a new library with re-encoded files, but in the meantime I'm experimenting a little bit.
I have re-encoded some albums to FLAC while applying DSP (cross feed and volume leveling). The files seem fine on playback, so I believe I can use this method if there are no other options.
However, out of curiosity, I analyzed the resulting files and discovered a couple of things I don't quite understand:
1. The dynamic range (DR) has changed, on one album typically up from about 8 to 10. Shouldn't the dynamic range remain the same, or, if anything, perhaps be a little bit reduced instead of increased, since the volume has been lowered? The R128 values are approximately unchanged, some tracks up a few decimals, and some down. What is the difference between R128 and DR here?
2. The new, volume leveled tracks have new replay gain values. In this case they have changed from around -7 on the original tracks to around +9 on the processed tracks. I would have expected the processed tracks to have replay gain values of around 0. Why is this not the case?