The "free dragnet to paid targeted forensics" framing is the right axis, and it maps cleanly onto what happened on the enforcement side. The 2024 wave taught everyone that the fragile part of CoinJoin was the party you could subpoena. JoinMarket survived because there is no one to serve papers to. But passive chain analysis never needed a subpoena, and maker change peel chains are exactly the structure clustering heuristics were built for: peel chains are the oldest primitive in the chain-analysis playbook, so a maker respending predictable change round after round hands the analyst their strongest tool back.
The underrated part of your writeup is the taker-side effect of option 1. A crowd of maker LN swap spends gives takers cover, but it also makes taker privacy partially dependent on maker Lightning hygiene: node pubkey reuse, channel announcements, and swap timing all sit outside the taker's control. Does the tiers-that-compose plan treat that as an accepted dependency, or is there a version where taker cover degrades gracefully when a maker runs their Lightning side badly?
The "free dragnet to paid targeted forensics" framing is the right axis, and it maps cleanly onto what happened on the enforcement side. The 2024 wave taught everyone that the fragile part of CoinJoin was the party you could subpoena. JoinMarket survived because there is no one to serve papers to. But passive chain analysis never needed a subpoena, and maker change peel chains are exactly the structure clustering heuristics were built for: peel chains are the oldest primitive in the chain-analysis playbook, so a maker respending predictable change round after round hands the analyst their strongest tool back.