Separate pools could be the way out: take the division by country and subscription status, but further divide each bucket by utilization. Ten percentile blocks for example. It would not have to be perfect to be well good enough. And I suspect that it would be easier to communicate in negotiations than an accurate formula, similar to how many countries seem to have ended up with discrete tax brackets instead of one universal formula across the entire range.
(And that doesn't mean that they could keep the data indefinitely or use it for other purposes without consent.)
This probably gets more hairy when it comes to a user's listening history and things like Spotify Wrapped, discover weekly etc., but this is one of those areas that fall under expectations from a service, legitimate interest etc.
This is solved (well, not solved, but handled) by not having a direct link between usage data and the user. So you store your data as <usage, proxy_id> which references <proxy_id, user_id> which references <user_id, user PII>. This way you can always easily break the connection to PII (if the user requests deletion) while keeping per user data.
But true, it's not anonymized in this case.
It used to be the case that the big record companies demanded pro-rata licensing. Tidal (before Jay-Z bought it) commissioned studies on user centric licensing and approached the record companies about it, but they were uninterested.
Now, however, it seems the record companies want user centric licensing, but Spotify resists it. It's hard to say exactly what kind of politics lie behind that opposition. I'll note, however, that Spotify seems to have adapted a lot to "gourmand" listening with their background playlists.