That's a tradeoff that could be made, but the onboard ion thrusters are about 5x as efficient as the rocket. They'd probably be much better off making the onboard tanks a bit larger and launching to the same low orbit.

I'd guess it is more efficient to have them boost themselves, so that they act as an additional rocket stage. (and so not need the delta V to elevate and then de-elevate starship's orbit.)

There is certainly someone doing the optimization problem for the orbit insertion and fuel loads and picking the right set of tradeoffs.

The optimization problem also includes drag losses at the lower altitudes.

And those drag losses are a lot lower keeping all the satellites inside the rocket rather than having each experience its own drag.

Might be worth putting an ion drive on the f9 2nd stage for that reason...

Re-entry is a bit of a feature, though - should a satellite be DOA, it will be in a low orbit and burn up faster.

I wonder if they could use another satellite to sweep a DOA satellite. They all start in almost identical orbits, so unlike most orbital degree cleanup plans there aren't huge relative velocities to deal with or anything...

It would mean that 1 DOA satellite means bringing down 2, but that might be worth it if the DOA rate is low.

Absolutely, but for the ones that are not DOA, having more fuel on board means they can re-boot themselves for a longer time, thus staying up for longer and getting a longer service life per sat...