Archival is a weird animal that doesn't behave like other things in tech. It's like scrapbooking. It doesn't have to be perfectly organized or available online. The archivist just has to be extremely consistent throughout his or her life, in order to make the files last for 50+ years, to be passed down to another extremely consistent person, and so on. Imagine WWI photos. In 1920, why would anyone keep those around? They had no message to portray, it was just a horrible recent event that you'd be called weird for trying to keep. Yet the few people in the chain of ownership were important for piecing together the photographic history we know today.

MIDI files are somewhat similar. Only a small population care, an even smaller population care a lot, but archivalists shouldn't be the one to determine to keep them around for 200 years. They just should, and forget it.

So yeah, that's pretty much what I do.

How's the quality of the data, is there any other cleaning that could be done / crowdsourced in order to improve the dataset?

An option to download the archive would be great, just because it's always good to have an offline copy. Any chance you could also add the ability to upload new MIDIs to the collection?

Just a thought: are there any potential copyright issues with this? I vaguely remember MIDI sites running into such issues "back in the day".