[htpcero]

Great, thanks for confirming. Your data has some wild and wacky artist and album "names", but does make for interesting testing. Thanks for motivating me to make a more robust effort at the stored functions that the music db has on MySQL/MariaDB to simulate the original in local memory sorting approach.

Yeap, re-migration from v18.X to v19.1 (skipping v19.0) would fix, also I can let the few (if you are not totally unique) with v19.0 on client/server and wacky album names have the SQL to run that will update the stored functions as needed. So I think that is the way to go rather than migrate everyone to a new DB that has exactly the same schema as before. I'll get a PR done and ping you when there is a test build to try, thanks in advance.

Meanwhile looking at your data....
I understand that you are working with historic cuesheets etc., but your artists node must look horrible and unusable. I did wonder if adding "," as an artist separator in advancedsettings.xml before scanning might clean it up a bit. There are a few artists with comma in the name e.g. "Earth, Wind & Fire" that would be mis-interpreted so not a standard or good choice of separator, but things like "Orchestra del Teatro dell'Opera di Roma, Leontyne Price (Aida), Jon Vickers (Radamès), Robert Merril (Amonasro), Plinio Clabassi (Il Re), Rita Gorr (Amneris), Giorgio Tozzi (Ramfis, capo dei sacerdoti), Franco Ricciardi (un messagero), Mietta Sighele (Gran" and "English Baroque Soloists and Monteverdi Choir, Furio Zanasi (Ulisse), Lucile Richardot (Penelope), Francisco Fernández-Rueda (Eumete), Hana Blažíková (Minerva/Fortuna), Krystian Adam (Telemaco), Gianluca Buratto (Tempo/Nettuno/Antinoo), Michał Czerniawski" as single records just make me cry. Then again you have used <lastname>, <initial> a few times and that would take the initial as an individual artist. Maybe bulk edit the cuesheets replacing "," with ";"? I'll leave both as suggestions.

Embedded metadata could do so much better.

you are right, what can I say
I do have a plan to migrate to embedded metadata at one point, I even have a working prototype script to cut the flacs into songs based on cuesheet and semi-automate the tagging. I estimate will take 2-3 days to run over the entire collection. If you notice, I tried to use some level of consistency within the Genres. For classical, I "overloaded" the PERFORMER field like so: COMPOSER; ORCHESTRA, SOLOIST/s; CONDUCTOR and this works OK 80% of the time. For non-classical, it's a lot simpler. The extreme cases tend to happen for Operas where I have COMPOSER; ORCHESTRA, CHOIR, CAST; CONDUCTOR. 

So instead of mass replacement of ',' with ';' my plan is to use a python script and code different tagging logic based on Genre and use it at the same time I cut each song from the album. What's keeping me from starting this is the daunting challenge that comes after - to fix the 20% of the albums that I expect will go not-so-well with the tagging logic and will need me to manually edit the tags. 20% of a lot of albums, it's still going to be a lot of albums to retouch

In the meantime, for Classical and Opera, I mainly use "advanced playlists" to slice my collection. I have advanced playlists per composer (that match the composer name in the performer "field" and also for top conductors and top soloists.

In the meantime thanks to the pandemic and I have had the chance to listen to my collection almost entirely in 15 months, again using advanced album playlists that select albums that are below a playcount threshold. All in-between work meetings . This had made the pandemic isolation much easier to handle.