I've heard it suggested that Starlink is used by SpaceX to keep their schedule fully booked, which helps them maintain quality standards, training, etc. If so, they could perhaps lose money on Starlink but still come out ahead when you factor in the benefit of having experienced workers for whom building and launching rockets is routine.
That's a great point. Similarly it's also a great opportunity for them to push the limits of reusability. A rocket failure with a paying customer's payload is Very Bad. But the risk tradeoffs look a little different if it's your own payload and you already have an assembly line churning out replacements
