> * Why did you remove the gh#627 reference?  Not relevant anymore?

I think the note about gh#627 warned the reader that a "vector to list
conversion" was necessary.  However, mapcar can work with vectors, so
the reference is not relevant anymore.

> * Could eglot--lambda be used for simplicity (I know it wasn't used
> before)

I think it won't be simpler, but it might be easier to understand the
intent of the code.  However, I don't know how it is possible to rewrite
the lambda.  How can we access the original object from the body of
eglot--lambda?

The patch has this: (lambda (a) (cons (plist-get a :title) a))

It should be turned into something like this:

(eglot--lambda ((MessageActionItem) title)
   (cons title orignal-object))

> * Are you 100% we'll return an 'equal' object for servers that don't
> make use of this feature?

The specification says: "result: the selected MessageActionItem".  If
the server does not send additional properties, the new version will
return what the original one returns: an object with a single title
property.  So this patch should not cause any trouble to servers
following the specification.

Although, there is a corner case when the new version gives different
result.  The original version can return (:title "OK") for an empty
action list, which is not in line with the specification.  The updated
patch returns :null in this case.

> * In the original version it was obvious the structure returned, now
> not so much.   Maybe
> a very short comment showing what the typical object looks like.

I added a comment and extended the docstring.

> After evaluating this, and doing changes (if any), feel free to push.

After your last email, I requested inclusion for the emacs group at
savannah three weeks ago.  Nothing happened since then, which I
interpret as polite rejection.

I've attached the updated patch.

Thanks again,
Felicián