Their “simplest possible solution” has not solved the problem.
Legit users who had modded their clients to have some perfectly fine features (such as performance enhancements) are SOL - they aren't gonna risk it just for features they've lost, they either gonna live with whatever's left or just abandon it entirely.
And those with malicious intent are gonna just work around the anticheat (because why not - what they're doing is most likely was some bannable offense already) and continue. Yes, there's some barrier but just the other day we had a thread where DMA hacks were mentioned (and PCIe cards for that are pretty cheap) - does EAC has anything against those tools?
We don’t put locks on our doors because they’re unbreakable but because they are the right compromise between cost and benefit
It should not require much extra developer time at all to let both users and creators choose whether to join an EAC or non-EAC instance.
I pretty much agree with all your points, but what I see here is that VRChat developers or product people are choosing the solution that requires the least effort. They may have other features or fixes in their roadmap and decided to squash all the fixes into a single one: banning them.
If I were them, given the popularity of modified clients I'd try to turn that into a feature that sets them appart from other competitors, but I still think it's a rational choice to ban them given how much developer time costs.