> I'm not arguing against collisions becoming more likely. I'm arguing against it becoming commonplace to the point that it becomes a commercial concern.

Minimising collision risk already is a commercial concern, and the number of conjunction avoidance manoeuvres SpaceX takes in order to achieve this has been growing exponentially (which presumably is a major factor driving their move of 4k satellites to a lower orbit which involves more station keeping) Obviously this gets harder when most of the satellites avoiding their orbits coming too close don't have the same owner, particularly if some of the other megaconstellations aren't even particularly cooperative (hi China!)

> Nobody is plane changing out of a collision. And for the foreseeable future, in LEO, the birds are not propellant constrained. (And launch is getting cheaper.)

No which is why I mentioned the fact that constellations pre-emptively plane change to avoid conjunctions. The frequency with which they have to do this scales superlinearly with the number of satellites operating in or intersecting the orbital plane. Ultimately propellant use for those manoeuvres and station keeping defines the satellite lifetime: agree it's not a huge problem when a satellite is only making small orbital changes a handful of times a year and its got a decent sized delta-v budget for station keeping and EoL deorbiting anyway, but another 70k satellites in the same plane would require quite a lot more adjustments, never mind them operating at aircraft density as proposed earlier.

> We're decades away from this being a problem. That gives ample runtime to developing e.g. magnetic station-keeping (if we go reactionless) or more-efficient engines.

Depends how fast the satellites get put up there (and also whether orbital megastructures become a reality, although non-trivial numbers of them actually might be decades away). There's some scope to improve propulsive efficiency (hi colleagues!), but within the power/mass constraints of a smallsat, you're not likely to see orders of magnitude more improvement in specific impulse over current gen EP, and we are forecast to need orders of magnitude more avoidance manoeuvres, which is generally going to mean more reaction mass. Sure, if we get reactionless propulsion suited for precise orbital changes in LEO then we can forget all about the tyranny of the rocket equation, but hey, if we perfect flying cars we won't have to think about the implications of congestion on the roads!