I just ran a test of my own:
Created 100 000 zero-length mp3 files each tagged with a 1-character Artist field. MP3Tag 2.46a read them all in just a few minutes - column sort takes about 12 secs. No problems with scrolling, just a little sluggish.
Test edit of file number 100 000 wrote to the correct file, with no 64K wraparound - the danger here was that if 16-bit values had been used for indexing the file list, tags might have been written to the wrong file, but all seems OK!
Looks like we're good to keep on ripping for a while! 