ˈt͡sɛːzaɐ̯: ya... 70k is still a lot. If it takes a minute to do a provide (that's what I think things are at for now although the performance should increase as more nodes update, secio is dropped, etc.) then you need 70k minutes of providing within the churn interval (let's pretend it's a day). That means having ~50 provides happening concurrently.

That's not impossible, but it's definitely not happening by default. You could turn off providing and have a script attempt that, but it seems quite excessive.

I'm not sure about your use case but if your application involved people naturally finding others with some similar content then using the DHT for content routing here might not be necessary. It may also be that the data needs fewer "hooks" to reprovide, e.g. folders, since your users tend to download many things anyway and not just individual files.

Solving this problem more generally is hard and is something the team is interested in looking into later this year and into next year. It has a ton of complexity associated with it though so we'll need to think it through.