I tend to agree it's not cool for the biggest client to just invent new standards without asking, but if he manages to make something that works and is backwards-compatible that could be good.

But I think the biggest issue here is that you've implemented NIP-94 on Amethyst in a way no one (I think) expected it to be implemented: for inline images -- which is probably a fault of the NIP and the discussions not ever being clear about their intents and purposes. We should address that by rewording the NIP.