[orbital-decay]

>When the satellite fails, it quickly deorbits by itself.

This also means that failure recovery will be quite tricky if possible at all. There are some downsides to other points too: such a satellite would work at very thin margins due to the thruster being inefficient with air as a propellant. Its ground swath width will be lower, coverage will be worse, requiring more ground stations (remote sensing is very often limited by the downlink bandwidth). Also, some kind of aerodynamic shape will be required, limiting its capabilities and power budget. (electric propulsion needs a lot of power itself)

[jccooper]

"Quickly" in this context is probably still weeks, and you could carry a little backup system to kick it into higher orbit in case of trouble. But really, low-flying com or imaging sats are probably parts of large, "cheap" constellations and meant to be of limited lifespan.

[alex_duf]

>This also means that failure recovery will be quite tricky if possible at all

Nowadays it's probably cheaper to send a new one than doing a whole Hubble like hot fix with a space shuttle

[orbital-decay]

I'm not talking about on-orbit servicing though. Failure recovery is done all the time with most satellites. The operator just needs some time to determine the nature of the problem, possibly doing some workaround. With a low flying satellite you don't have much time for that.

[walrus01]

failure recovery in the context of spacecraft usually means software-commands sent via TT&C (tracking, telemetry and control) channel to switch to another piece of hardware, part of the N+1 or 1+1 configuration on the spacecraft. It is incredibly exceptionally rare for a human to ever visit a satellite once in orbit.

They did it a few times in the 1980s with the shuttle, including recovery of a satellite to prove it could be done, and there were the hubble servicing missions. But other than that no human has ever touched a satellite once it's in orbit.