Ask an ye shall receive. Warning: Tall weeds ahead. I may be totally wrong on some or all of this.
As already said the MKS is faster .03sec/60deg vs .05sec/60deg. Will you actually notice that? Probably not. The hardest, craziest smack pilot wouldn't be able to notice it. 2 hundredths of a second is nothing.
As with everything there is a trade off. While the MKS is faster, it has less torque: 48.6oz/in vs 69.4oz/in. Will you notice that. Possibly. If it were on a cyclic servo, definitely. On the tail, it is less critical. This will translate into less power in the hard starts and stops of tail movement (fast piros from stable flight/hovers or fast stops/reversals from fast piros). The torque forces on a tail servo are significantly less than on cyclic servos, though, so the actual difference may be too small to notice.
So, why do they have these super fast sweep time, fast signaling servos for the tail?
First, what do the numbers mean? The MKS's 760us/560hz timing info indicates that it has a 760us center pulse width and can operate at up to 560hz refresh rate (new position information showing up at the servo 560 times per second). The lower the center pulse (1520us or 760us are the two you will see) means it has a shorter overall pulse timing and, when coupled with a faster refresh rate, allows for the servo to be, theoretically, more precise and allows for lower latency in the servo seeing a new position input value and making it happen on the output shaft.
Remember, the servo has a PID controller inside it, just like the FBL or gyro unit has one on each axis of control. Where the FBL/Gyro is using a PID controller to maintain the rate of rotation on each axis as set by your stick inputs on the Tx, the servo's PID controller is working to put the output shaft at the position indicated by the input signal at all times. In order for the PID controller to have more precision and lower time between input change and output position settling, you have to increase it's cycle time (number of iterations through it's calculations per second). But this speed up is also dependent on the input value to the servo's PID, which is provided by an analog to digital conversion process (convert the pulse train into a value between 0 and 2048). If you want faster conversions you have to speed up the signalling rate and modify the electronics to filter appropriately for the given signaling specs. All of this combines into a more precise output control, and faster positioning of the output shaft on changes in the input signal value.
So the faster timing (760/560 vs 1520/333) is important because it allows for more precisely maintaining the position of the servo (better centering, better hold at a shaft position, etc.) and because it allows for shorter latency between the input changing and the shaft starting to move and ultimately achieve it's new stable position.
Specs almost never include latency information, though.
The servo's speed is used as a sort of stand in for the latency because there is an assumption that the testing method used to get the sweep time is based on the stopwatch starting when the new position input is asserted and the watch is stopped when the shaft is at the new position. In fact this is not generally the case. Sweep times are usually generated by putting the servo at one end stop, sending a signal to put it at the other end stop, and then the sweep time is grabbed at a 60 degree arc in the middle of the overall sweep it is executing (either by using high speed camera, an optical start/stop trigger, or an electrical contact start/stop trigger). This way the spec for sweep time does not include the time it takes to start and stop the gear train, nor does it include the latency in processing the new signal and beginning the move, it also takes the PID control's logic out of it (it is controlling over a large change in position so it's logic whittles down to more or less full power to the motor during the part of the sweep that is captured for the spec).
Not long ago there was a video or two put out by someone, can't remember, that did a sweep test in the manner lay people expect: start the watch when the new signal is asserted to the servo and stop it when the movement of the servo stops (indicated by low current draw). The testing was done on a couple of servos using the servo test feature of an iCharger I think. I don't recall the "real world" sweep timing vs the manufacturer provided timing, or what servos were tested. If you are interested might want to do a search for the vids. In general I'd expect a 10-20% increase in sweep time spec using this method of testing vs the in-motion sample method usually used by manufacturers.
And now the last bit, talking more about the timing, relationship between center pulse and refresh rate, etc.: There is a relationship between center pulse and frequency. The maximum frequency you can drive a given center pulse timing at is restricted to 2x the center pulse period. So: On a 1520us (microseconds) center pulse, that's 3040us total pulse width which is equivalent to a 329hz signal (1 / .00304sec). There is a small area on either side of the signal that is "ignored" and not part of the usable duty cycle of the signal pulses so they can run at 333hz safely, but not much higher. At 760us center pulse (ie. half pulse width) the maximum drive frequency is in the neighborhood of 658hz. Unlike with the 1520 pulse signals where they overdrove the refresh rate to 333hz, they will not be able to get away with overdriving the frequency on these signals as much due to the smaller buffer zone on either end. 660hz is probably the most they will ever overdrive it, assuming there is ever a need to go above the current 560hz refresh rates (I doubt it).
LOL