> so why can't we simply retract them so that enough dust falls off?

Thus scratching the delicate surfaces of the panels, as well as requiring a retraction motor and associated devices that have their own failure modes.

I don't buy the "solar panels can be scratched" argument when they're routinely surviving dust storms much more violent than that. I also don't buy the "you need a motor et. al. for retraction" when the one that already exists for extension could likely also be used for retraction - or at least enough of it to give the panels a nice shake.

What I do buy is that panelshaking of this sort uses energy that would be in short supply if the panels are sufficiently blocked to require it. That doesn't seem impossible to solve, either, but I don't blame NASA for not feeling the need to solve it when they already designed and built InSight to exceed its mission parameters.

I believe the answer is that a brushing mechanism would scrape the glass, and some of the dust gets adhered due to static electricity. I've also heard speculation that diminishing returns are a significant concern when it comes to a dust removal mechanism because of the reasons above.