Most of the time this is handwaved over, for bunches of reasons. Part of it is that there is no one valid frame of reference: you can't just say "this stays stationary" - because what does that even mean? Stationary relative to what?
If you mean "you go wherever inertia would take you in that amount of time, with no acceleration", then if you go backwards a multiple of a year you end up pretty close to Earth (quick approximation: the Sun's acceleration in its orbit in the milky way is ~2*10^-10m/s/s, which works out to ~100km in a year, ~400km in two years, etc.). In actuality you could probably get somewhat better range by letting your velocity build up to intercept Earth-that-was at the target time. (You'd need to take into account a bunch of things for best accuracy: the Earth's orbit is not "perfect", the moon's phase, the acceleration of the Sun, etc, but we know most of that.)
You probably would end up needing to make multiple jumps, pausing in-between to let Earth's (or another planet's, or even the Sun's) gravity well readjust your velocity to what is needed for your next jump.
This means that you'd need to build in re-entry capability into the machine (as you are unlikely to have the precision necessary to be able to jump into atmosphere at a sane velocity), and potentially maneuvering delta-v. Note that with a time machine you lose conservation of momentum, which makes things a lot easier. You can change the direction of your velocity vector by jumping back in time and interacting with your past self. You can potentially use this to get "free" (read: reactionless) velocity out of any gravity well.