Comment by jcrawfordor

when you're trying to take relative measurements of the motion of objects within the field of view, which is how you do these fine position measurements, you don't have a lot of choices for objects with a measured location accuracy similar to the measurements you're taking. You're measuring the precise position of something compared to other things you don't know the precise position of. That requires a very stable platform so you can do comparisons internally. There are lots of options for precisely determining where you're pointing but pretty much all of them involve a loss of precision whenever the platform moves, so if you have to slew to a guidestar and back your accuracy is limited to the measurement of the guidestar and then the error induced when you slewed. You also end up using different instruments for alignment measurements vs. the actual observation (for example because the observation requires a long time at a static position), and there's an imprecision involved in comparing the different instruments because they slightly move relative to each other with thermal effects and so on. When you really get into it, using some instruments will reduce the precision of other instruments because they vibrate the platform or create heat. You have to account for all of this with a complex model.

In practice these highly precise measurements, at least in the domain I'm familiar with, become sensor fusion problems where you take a lot of sources of position info, weight them based on their accuracy, and integrate them over time. The less stable the platform, the more error is induced by the integrating over time. Nothing in that realm is really all-or-nothing, as we're seeing with Hubble as it racks up more and more failures, but the loss of the rotation will mean more error in combining position references which will mean less accurate final observations. They may no longer be that much more accurate than measurements obtained by other means.

I'm not sure if I explained that very coherently, it's a complex field that I used to write software in but, well, I was the person writing the software, not the person figuring out the theory. The general idea is that space-based instruments tend to have a bunch of different factors that go into their final accuracy and that accuracy normally gets worse over time as you run out of fuel and things degrade and ultimately stop functioning. Fortunately since space-based systems cost so much to build and launch, the teams behind them have usually put a lot of thought into how they'll continue to get the best use out of them as they get older. That often means having future plans for different missions that just don't require as much accuracy, which is the case with Gaia---it's ending this "phase" of the mission plan.