

Moreover: changes that big go in the next game.
If you decide to roll your own engine, from the start… awesome. Especially when content-creation tools aren’t a huge deal, like if your game world is procedurally generated. Understand your scope, steal freely from existing libraries, avoid wasting your artists’ time.
If you try to switch engines mid-development, you are fucked. John Romero couldn’t make that shit work, just going from Quake to Quake II. Daikatana could have shipped before Half-Life and only missed colored lighting. Instead it’s a cautionary tale. Duke Nukem Forever didn’t even ship. They had a nearly-complete game, several times, but threw those out and started over.
You don’t have to throw anything out, to start a new project from scratch. Ship the damn game. Put different tech in the next one. If you don’t ship, there won’t be a next one.
A game with that title shipped. I don’t think it was, in any sense, the same project. Not even as a ship-of-Theseus situation. A different studio started from scratch.