Been digging into JoinMarket maker clustering on mainnet. the short version: yes you can cluster makers by fee fingerprint from onchain data alone, and yes it reduces taker anonymity sets. But JM holds up pretty well in practice. mean anonset goes 7.6 -> 6.9, and the mitigation (fee policy homogenization) *can* be a client default change, no protocol surgery needed. The counterintuitive part: some "obvious" countermeasures like makers avoiding change as input actually make things worse. Rough draft, not peer reviewed, happy to get feedback:

JoinMarket Maker Clustering and Taker Anonymity-Set Reduction