One of those arcseconds passes by approximately 15 times a second, so using an atomic clock for that is way overkill - any time source more accurate than 1/15 of a second is unnecessary.
The main problem is that once you have aligned your telescope, now what you are trying to do is measure the direction of gravity, compared to the direction you're pointing the telescope, and that's a lot harder. Partly because that's a mechanical angle measurement (you need a pendulum that can freely swing and rest with minimal friction pointing directly in the direction of gravity, and then you need to measure angle). But then also, you need to take account of the fact that the gravitational field on the Earth is lumpy. If you're sitting next to a mountain, the gravitational field will be deflected slightly from pointing directly downwards - and in fact that was used a while back to measure the density of the Earth.
But theoretically, if you can solve the angle measurement bit, you could determine your location on Earth with an error of around 30m.