Another interesting example is a story by singer and Gershwin scholar Michael Feinstein. When he was a teen, he fell in love with George Gershwin's music, and discovered that his brother, Ira, the legendary lyricist who supplied the words to most of George's songs, lived nearby.

He went to Ira's house and knocked on the door, and introduced himself. Ira was happy to talk to the kid about he and George's old songs. During the conversation, Ira opened up the piano bench, and it was filled with old manuscripts in George's hand of totally unknown songs that had never been published.

Feinstein ended up being the annointed by Ira as the unofficial Gershwin scholar, and he later recorded many of those unknown songs.

He also told this story on NPR's Fresh Air:

In 1982, there turned up in Secaucus, N.J., at the Warner Brothers Music Warehouse, which is the place where Warner’s kept all of their stock of their published music that they would sell. Suddenly somebody who was working in that warehouse, a guy named Henry Cohen (ph), found these boxes and boxes of music that looked like manuscript material of not only George Gershwin but of Victor Herbert and Leonard Bernstein and Arthur Schwartz and Cole Porter and Vincent Youmans and on and on and on. And we were called - we, being Ira Gershwin, for whom I worked in 1982 - he was - what? - 83 or 84 at that point. And they said, there are these manuscripts of George's here and lyric sheets of yours, and somebody better come and look at them. And Ira said, oh, no, that stuff was destroyed long ago. There’s - that’s a mistake. So he sent me to look and see what was there only because of the insistence of the folks in New Jersey, even though Ira was convinced that we would find nothing. And it turns out that I found, amongst all these boxes, 87 original manuscripts in George Gershwin's hand plus copies of scores that had been lost for 50 and 60 years. For some reason, they were all there, and it turned out to be one of the greatest musical theater discoveries of the century.

Yes, for sure. But then that's a "lost cache". Similar to the works we'll find in a few years buried somewhere under the ground. But what about "active collections"?

For sure I would also expect that the division line lies somewhere else than just one continent. But so far that's the only datapoint I have