It’s easy for design to call for things that preclude some implementation options. And the ones that are left might preclude other design choices you’ve made. So I think it’s generally relevant for the conceptual and user experience design to be in an active feedback loop with high-level ideas about implementation.
But, of course, it comes down to details. If I wanted to do some super-simple take on a CYOA game that had no need to even maintain state, well, I already know HTML and could just write it in that far faster than it would take me to learn an existing framework (any of which, of course, would support much more than the super-simple take on CYOA I’m positing). I wouldn’t suffer for having ignored thinking about implementation details.
If I wanted to write a parser-based game with a parser and world model that went way beyond what Inform 7 or TADS or Dialog facilitates then, well, if I won the lottery and quit my job and worked on it full time then with reasonable luck I might have something to show in two years. In a world where I’m not likely to win the lottery, that comes to… something I would never finish. And after a decade or so I’d probably wish I’d given some thought to implementation.