Well i think the right way from simulation perspective, would be to add acceleration the duration of active truster. For agile ness this acceleration coul be powerfull enough to be short enough to just smooth out the movement. With a little over schoot by the deceleration thrust. A counter puls of deceleration.
The bigger the ship the power mass ratio is lesser so slower rotation but also longer overschoot and very smooth respons. This give the feel of flying something massive or big.
You get the effect of short thuster pulses when light fighter is moving a lot dodging enemy fire.
But big heavy salverger has longer burst but much weaker compared to the mass it needs to turn.
It's basically a question between having a slightly underdamped response (which will overshoot and then settle down to the target speed via counter thrust) or a critically damped response (which will hit the exact target speed with no overshoot, but will also take a bit longer). To my knowledge (based upon the interview with John Pritchett), the IFCS is currently set up as critically damped, however they might have changed this with the introduction of jerk.
You could of course also have an overdamped response, but that wouldn't really make sense from a realism perspective (a critically damped response would always be strictly better), although it might make sense from a gameplay perspective (an overdamped response would basically take significantly longer to hit the target speed, than both underdamped and critically damped responses, thus making the ship less maneuverable and with a greater feeling of inertia).
The mass of the ship doesn't really matter as such in this regard, as the IFCS shouldn't have any problems adjusting for it. Whether you overshoot or not is entirely down to how fast you want to hit your target speed and then tweaking the response correspondingly. Of course, the total length of the thruster burn would be longer with more massive ships, and as such it might be more attractive to balance your thruster response towards an underdamped response (which is faster, but overshoots), but again it's a question of preference.