Atmospheric distortions are worse when looking horizontally though the atmosphere (think marriage on the horizon).
It's also only the bottom few miles of atmosphere that is really thick.
Looking vertically from 400km up, there actually isn't too much distortion. The limiting factor is making high enough quality lenses and getting enough light onto your sensor without too much diffraction.
The previous post argued that the influence of the air is less severe looking straight down. If we now go back to looking at an angle, we are back to square one, the problem of distortion due to the amount of air between the observer and the observed.
If I'm running the numbers right, then using the Hubble's maximum theoretic angular resolution of 2.8e-7 rad[1], something 400km up would still not resolve features smaller than 11.2cm. That's the absolute best, diffraction-limited case. Don't know how close we are to that on real-world optics.
There's probably a way to circumvent these limit by using multiple satellites working together[1], combining images to form a synthetic aperture[2], using an aperture mask[3], or a combination of any or all of those. Actually I'd be surprised if state-sponsored spy agencies were not already using this.
I forgot about synthetic apertures! IIRC the Event Horizon Telescope result had to include the model of what the black hole probably looked like to refine the image, it would be fascinating to see what could be done by a constellation of telescopes in LEO. Though the [3] link says that aperture masking only gets you to the diffraction-limited regime.
It's also only the bottom few miles of atmosphere that is really thick.
Looking vertically from 400km up, there actually isn't too much distortion. The limiting factor is making high enough quality lenses and getting enough light onto your sensor without too much diffraction.