Might be unenforceable, but they don't have to cater to them and can and will break forked clients for however long it takes those to catch back up. And then Signal gets flak for not hobbling their development processes to the speed of the slowest (somewhat popular) fork.
This doesn't have much value without federation, which would require active support from OWS.
You can literally just look at Matrix, the tool you recommended, to see the problem with federation, fragmentation, and market demands to support lowest-common-denominator security. It took years after your recommendation for Matrix to have universal default E2E, the table-stakes feature of a secure messenger.
Are you going to back this up with evidence, or is it okay when you use axiomatic arguments?
"It took a while" is not actually an argument that the procedure is wrong or less correct, in case you were unsure.
In fact, it is an argument that you were wrong, and the evidence for that is that when you made the argument, and for two whole years afterwards, your recommendation would have put vulnerable users on a platform that was for many users default-plaintext.
There is, of course, the vague language of the CFAA, so I’m not sure I’d want to test this theory, but his demands that forks not use the main centralized servers are, in my opinion, unenforceable bluster.