To leverage this, it's fairly easy to detect you're on a self-hosted VPN: your IP is in an IP range assigned to a hosting/colocation provider, is not a TOR proxy (there is a public list of those) and doesn't belong to any remotely popular VPN (easy to enumerate for a little money, lots of lists exist).

In exchange for that you have eliminated your ISP (or public wifi) as a threat but instead added the hosting provider to the list of threads. And for any adversary that stands above the law, the routing infrastructure of your hosting provider is already a valuable target.

My Streisand hosted on AWS looks to the outside like anybody else's Streisand hosted on AWS, doesn't it?

Similarly, my f-secure egress looks like anybody else's f-secure egress, so what's the difference?

I don't really know, I don't use a VPN. Really asking.

They certainly let law enforcement and intelligence agencies know, often without a warrant.

Please read my comment as if the threat model includes panopticon governments, not common skids running aircrack-ng.

(I'm not saying you're wrong; again I've not really thought about having to thoroughly anonymize my own traffic.)