This industry has a serious problem in a few areas:
1) Lack of controls during the prototyping pre-alpha stage - LOCK DOWN scope early, define the project early, and then allow your leads to have the autonomy to make micro-decisions within the overall scope you have established.
2) Scope screep. Completely gut superfluous items in the game that add nothing, bog down development, and only add potential for more things to check, verify, de-bug, etc.
It amazes me that we can't get to that model yet. It should even work well with new IP. Smaller teams get a chance to play around with mechanics and then once they get something that works, fully flesh out the concept of all the surrounding systems before marching head first.
I feel like some games have none of that planning - Starfield being an egregious example. Other games seem to be far more efficiently planned (Square's recent Final Fantasy games, FromSoft's titles, Capcom's games, Insomniac, etc. etc.). Sure, some of these games still take a long time, but they generally release in a very good state and I don't think the dev teams are churning through massive scope deviations so the production vales are super high, something that is severely lacking in something like Starfield.