OK. But instead of adding structural changes to eglot.el, how about describing exactly the use case that you're thinking of solving?

Earlier, I had the impression that your use case concerns usage of eglot-ensure in a specific scenario, which isn't perfectly described (as in a formal bug report that I myself can witness). So why not do one of these things, or both:

• describe the minimal reproducible example that shows evidence of the bug
• think about touching eglot-ensure?

When magit created a non-file buffer(such as magit-blob-previous on a git tracked file),
it will temporarily set buffer-file-name for the usage of normal-mode:

(let ((buffer-file-name magit-buffer-file-name)
          (after-change-major-mode-hook
           (remq 'global-diff-hl-mode-enable-in-buffers
                 after-change-major-mode-hook)))
      (normal-mode t))

Thereby eglot--managed-mode is enabled on that buffer.
Then when you move around in that buffer, eldoc is triggered, and there's error: (wrong-type-argument arrayp nil).

And if you think current changes are too structural, perhaps consider adding an option to toggle automanage? 6900927

it will temporarily set buffer-file-name for the usage of normal-mode:

Magic lying then, right? It says there is a supporting file, but there is none. I can't tell if this is the problem from your short description. I also have no idea what normal mode is.

Also your current implementation breaks all Eglot tests.

Adding a toggle to something as fundamental as Eglot's most basic working principle sounds more complicated than perhaps asking Magic matters why they do those things.

Well I agree that it's more of a magit problem, so I'm closing this.