I'm running 2.8 as I found a forum post about needing it for large numbers of tracks.  I am specifically working on Toscanini - RCA Complete Collection, which comes to 1035 tracks total.  It has crashed twice this afternoon as I have been working on this release.

The first crash occurred after dragging the files from the file browser to unclustered, then clustering them.  Upon scanning, Picard found the main release and some tracks that went to a couple of other releases (not too surprising with these reissue box sets).  As the file matching to the downloaded data was nearly done - somewhere around 850 - 950 tracks completed according the status bars, Picard crashed suddenly, with the window closing abruptly.  I think the log file was lost, as after restarting, only the startup debug output was present.

The second crash came after I followed a different procedure.  I dragged the files over and clustered them, but then I looked the release up in the browser, and triggered the data downloads from the internet browser window (Firefox).  I was using a simple renaming script to create sub-directories for each of the discs using the disc subtitle tag, which I was editing for each set of track on each disc.  I would select the disc tracks, edit the discsubtitle and then click elsewhere to have the subtitle tag saved (otherwise I had to save the files twice to get the correct renaming)  After each disc set, I'd click save and then remove.

After approximately 47 of 85 discs had been saved in that manner, Picard vanished while saving/moving/renaming the latest set of tracks.  I had the task manager open, and Picard's memory usage was between 200 and 300 MB, normally about 220 MB.  CPU usage would go to 25-30% on the Picard app while saving and then fall to near 0 while editing.  I had the manager open because Picard became unresponsive for several minutes when I dragged the tracks from the cluster (middle pane) to the update pane (on  the left).  There was no status update, and then suddenly all the track were in place. Also, I noticed a bug where there was very large memory consumption with large file sets, but I did not see that with the 2.8 version.

I've seen this sort of crash on the 2.7.3 version earlier, as well, but I'm not sure if it is related.  I worked around it by processing a smaller number of tracks at a time and then I found the forum post about increasing the data lookup capacity in 2.8.