I’ve proposed a solution for reducing spam on Global Feeds, such as Adaptive Proof of Work: https://observablehq.com/d/295d8a1c6f7b07f9

Perhaps there are simpler solutions but this one proposes using a Proportional-Integral-Derivative Controller (aka PID Controller) to tune the PoW required so that the rate of events stays under control. What it does is find the correct PoW required to maintain a reasonable rate of events (e.g. 1 event / 10 seconds).

PID Controllers are used to control robot limbs, dam gates, etc.

IMO it should only be applied to public keys that are detected to be abusing.

I was also proposing the idea of relays gossiping about spammers, once a relay considers an IP or pub key to be spamming, it sends an event including the pub key or hash of the IP and proof of the spam (merkle tree proof of event ids of the abuse) to other relays which can then be on alert. This didn’t get too much traction but I might implement something soon for Nostream relays… it will be opt-in of course.