Even before SpaceX started launching their own satellites in huge numbers they had a business model where they were selling the launch, not the rocket, and selling it at a fixed price, so if some small refinement saved them 5% on launch costs it went to their pockets so they had an incentive to make those small refinements.

Dragon 9 was based on conservative and boring technology but it was cost optimized before it was reusable, then reusability crushed the competition.

For that matter, Starship is boring. "Throw at the wall and see what sticks" isn't "trying a bunch of crazy stuff" but trying a bunch of low and medium risk things. For instance, development of the Space Shuttle thermal tiles was outrageously expensive and resulted in a system that was outrageously expensive to maintain. They couldn't change it because lives were at stake. With Starship they can build a thermal protection system which is 90% adequate and make little changes that get it up to 100% adequate and then look at optimizing weight, speed of reuse and all that. If some of them burn up it is just money since there won't be astronauts riding it until it is perfected.