My reading suggestion is, read this :) The reason a fidelity bond is a reasonable suggestion is that we need to impose some cost or scarcity to prevent an attacker creating 100k Sybil agents as potential coordinators. I'm assuming you don't use any kind of trust or public identity behind the process of advertising yourself as a potential coordinator. Why a fidelity bond vs some kind of more "high tech" thing like a podle commitment, a ring signature or an aut-ct token (think of it like a ring sig but you can get 1M anon set instead of 500-1k)? Because while privacy is important here, cost imposition should also be high. Something like aut-ct is good for rate limiting (I don't want a burst of 100k requests in 1 minute) with super-high privacy, but it doesn't impose any really significant cost on an individual user. A fidelity bond can lock money for a significant time e.g. 1 yr resulting in a very non-trivial time value of money sacrifice. While it isn't perfect privacy wise if on-chain, it can be an output from a coinjoin, so if well controlled, that utxo needn't leak a huge amount. Finally though, on the general idea of a "moving" coordinator: it ends up rather similar in spirit to what I implemented in Joinmarket with "directory nodes" to find lists of counterparties, then the taker pings the onion services of the makers, avoiding having the actual coinjoin coordination occurring at a central party. Here it could be seen as better, but I have two practical concerns with it: 1/ isn't a big part of the coordinator role to ban utxos that don't cooperate, if you had 1000 coordinators, or, let's say, the coordinator is always moving, is that mechanism impacted? and 2/ wouldn't you need some way (like "directory nodes") to discover the "current coordinator list"? hence "directory node" idea in Joinmarket (which tangentially ended up never really working properly due to inability to have the onion services accessible, I honestly still don't know how that could be fixed). Back to "which mechanism to prevent Sybil": none of them are perfect. With fidelity bonds you have to at least consider the problem that if you set the minimum cost to some reasonably-easy-to-attain for a normal person, you have the issue that a rich attacker can split their stash into 1000 smaller utxos and Sybil anyway. Chris' original solution to this was to use (sacrificed value)^2 but given the natural distribution of people's wealth, this *also* led to complaints about the fact that some people are 100 times richer and can therefore offer 10k more sacrificed value^2. It gets really tricky to handle this.