It is simple if you consider accelerations (both variable and constant) in "Special Relativity" as being made of piecemeal jumps in velocity. You can then apply Special Relativity to each velocity plateau where you may theoretically consider deferentially short periods of time where velocity is constant. You then sum up the small steps between one constant velocity and another different constant velocity (this is an acceleration) where a Special Relativistic System has been subject to acceleration (constant or otherwise). Special Relativity becomes an analytical form of General Relativity from the POV of a fixed "observer" frame of reference. Movements of clocks from one step in a high rise building to the next involves a change of acceleration due to "Gravity"... a small but measurable amount of "differential" time dilation is nowadays easily measurable.
According to the Einstein Convention you can only compare clocks in the same "mutual rest frame". If two frames are in relative motion then to set the zero on clocks you will need to bring one of the two frames into the chosen "rest frame". When clocks are moving relative to each other the "ticks" of their clocks are of different relative length but appear the same when considered from each other's frame of reference because "everything is relative". What is not accounted for is bringing one or other of these moving systems into the frame of the observer. There are two ways to do that, an observer in relative motion can either accelerate UP (or DOWN) to the speed of the relatively moving frame and then you may synchronize clocks, or you can decelerate or accelerate the moving (observed) frame UP (or DOWN) to the velocity of the observer frame. In the first case the observer undergoes an acceleration. In the latter case the observed frame undergoes the acceleration. Once this has occurred you use Einstein Synchronization light pulses, at least in theory. In practice this never really happens but please bear with me on this. In terms of General Relativity and the Equivalence Principle (the one on which GR is based and linked in Wikipedia below),
the systems are entirely distinguishable. It is not an arbitary decision as to how the clocks are brought into the same rest frame and synchronization in the one place and at relative rest. After that synchronization instant, the two respective clocks return to their respective moving frames. You me only one of these frames actually undergo acceleration (not "in fact" but only "on paper"). One or the other of the two frames and it's contents undergoes extreme acceleration. You then "assign" the same initial time to both systems. After that it is clear which system is undergoing time dilati0on and which one is not. Here is a full mathematical description of the case which is completely acceptable way in which Special Relativity and General Relativity agree. See:
How Do You Add Velocities in Special Relativity?
This reference is taken directly from John Baez's own thread. If you look you will find other sources following the direction of this treatment too. You can split the accelerations up into small steps (as per calculus) and the result is identical to the way you do it in General Relativity in the end, outcome is the same. You can use the fact that systems are instantaneously at rest to synchronize clocks and this is the way Special Relativity deals with Einstein Synchronization of clocks, it just does not state it explicitly. Consider that you bring to rest one of the two systems as they pass in close proximity to each other. Your choice of "rest frame" affects the zero time synchronization on the pair of clocks. Then semi-instantaneouly accelerate that temporary quasi-rest frame back up to speed (a significant fraction of the speed of light if you want to see a big jump) or through a series of incremental equal steps in velocity (a constant acceleration) or just one single BIG change in velocity. In the latter case that one step in velocity better not be the observer frame because it would crush to a fine powder every bone in your (observer) body, that is just the real effect of relativity - special or general on matter. Capiche?
It rams home
Einstein's Equivalence Principle and the way physicists like Einstein have chosen a very elegant means of solving the problem for highly symmetric systems
. So "Beauty" conquers "Truth" in this case. The not so beautiful is more instructive when you understand the "under-story" of Special Relativity. Unfortunately in the general case where we have an ideal gas where "billiard ball-like" particles are in relativistic motion all the time and changing velocity through mutual scattering, time dilation is a much more involved problem, tiny time dilations accumulate between "impacts" and frames are shuffled around (consider that as quantum entanglement shuffling about), and exact solutions using General Relativity become the more complex approach. This Special Relativistic Approach is "easier" provided you have a way to equate mass and acceleration, it is just that the "rest frame still must be chosen in order to determine which "twin clock" is to undergo time dilation and which one is the reference clock.
The way you should consider time should be understood in the context of a Page & Wootters Mechanism for those who have an interest. From that you can gain an "overview" of the real mechanisms behind the processes of time dilation through quantum entanglement frame changing. Check out this easy read reference:
Quantum Experiment Shows How Time ‘Emerges’ from Entanglement - Medium - Oct 13 - 2013.
And so we come "full circle".