I don’t think problems like this are totally unique to WaveLab.
For this reason, I always do a full render of the montage as one long floating point file at the same sample rate, and have WaveLab create a new montage from the resulting file so all the info is carried over (markers, CD-Text, etc.).
Then I can add a dither to 24-bit and safely render a WAV of each track, or SRC that full render down to 44.1k, use Custom Montage Duplicate, and have a full identical montage at 44.1k to render WAVs of each track from…or DDP, etc.