But if you're not gonna ship it to all your millions of users and receive the outrage as feedback, how will you ever know if it works or not?
A/B testing to a few users only works in web-app front-ends, not in professional tools where all single end-user releases need to look and act exactly the same.
Shipping to only a handful of users wasn't a thing in the era of "Gold CD" releases, and even in the era of the internet updates, nobody will want to take part in A/B testing and end up with a different Photoshop UX than what his colleague is using, so you either ship to all or none.
So it seems Photoshop's UI is more of a cause of inertia and resistance to change, rather than nailing right the first time.
A/B testing does very little to improve any UX. It's got merits in performance optimization, where the implementation differs but the contract is static between A and B, but with user interfaces, it generally leads to pessimizations as usage is not proportional to usefulness.
The rare exception is single-purpose interfaces where increasing one singular interaction is an end in itself, e.g. a marketing page, but that's a pretty unique case that is very far removed from a productivity tool.
AZERTY is very bad and France is still stuck with it, despite Québec having a variant of QWERTY for decades, ditto for Switzerland having a custom QWERTZ, and BÉPO being heads above.
Every time I'm confronted with an alphabetical keyboard my brain malfunctions. It should be easy but I'll be on R or S looking for the T and spend 5 minutes scanning for it. Usually via remote with some crappy app trying to login or search and its on a big TV huge, usually people watching making the awkwardness even worse since everyone else can obviously see where the T is.
I am not native but I feel pleasure from using Dvorak and chatting with no need to look at keyboard. I am trying to spread acknowledgement about Dvorak among young folks and those who agreed to learn feel the same boost. The difference is unimaginable for those who already has some muscle memory for anything related to Qwerty such as Ctrl-C. So the difference is between creating the layout for touchtyping goals and creating the layout for any other ones.
And sometimes there's good enough and you should leave it alone. QWERTY isn't optimal but it's not very far from it.