This is about bundling decisions together that do not need to be bundled. Changing the game engine after the fact is about the hardest thing you can do, virtually impossible after a certain stage. The decision for the engine is necessarily a very very early one.

But you can easily and very likely need to improve your backend many many times, even far into the games development or after release. Limiting your options for that so early is very risky.

And your are right, that risk could be offset by some large enough benefit you only get with that engine. I just fail to see where that would be. I can still use all of AWS if I choose an independent game engine.