My thinking is that it's very low-cost to support all possible encodings,
even if it's unlikely that a terminal would actually send them.
Using your example:
\e[65;5u
Would be a valid way to encode C-a according to the spec at
Most terminals will not encode it that way, and will instead send ^A,
but it would be nice to support it gracefully if a terminal happens to
send C-a encoded that way.
In addition, the "just support all inputs encoded this way" approach
seems simpler to understand and maintain than an approach that
distinguishes between key combinations that have an existing
alternative encoding and those that don't.
The entries in the keymap won't be referenced unless Emacs actually
receives matching input, so the cost of having entries for additional
combinations seems fairly minimal.
Maybe I'm misunderstanding keymap performance or some other
important parameter, though?