It's actually quite easy if you take the time and go through it slowly. Everything is pretty much simple math, just using terms that seem complicated. Chief Blur Buster explained why changing the strobe duty has an effect on crosstalk (you would think it shouldn't, but it does, and it's not desirable at all), but the important thing is the Strobe Phase setting.
The Strobe Phase has 100 values. These 100 values are positions on the screen that are cut into 100 points. Changing the strobephase changes the "timing" during a frame where the strobe begins and ends. The timing range is equal to the refresh rate frame time, which is pretty self explanatory: 16.7ms (total) at 60hz, 10.0ms at 100hz, 8.3ms at 120hz and 6.9ms at 144hz.
I think that's pretty easy to understand, right?
So, at 60hz for example, 16.7ms is split up into 100 "points". Divide 16.7ms by 100 and what do you get? 0.167ms.
So changing the strobe phase by 1 point changes the "time" on the screen where the strobe signal "begins", increasing it by 0.167ms per point. (you can think of it as moving it down, but screen refreshes are done top to bottom, so you're actually moving the strobe position "UP" (higher) up the screen as you increase it. That's why a very high strobe phase will have lower input lag than a strobe phase of "0". (Think about this for a moment and it fully makes sense, since the top of the screen (on any monitor, even non strobed monitors) will have lower input lag than the bottom).
How does strobe duty factor into this? Simple. Strobe duty is the PERSISTENCE--how long the backlight is flashed on during that single frame. Back to our 60hz example, each point of strobe duty adds 0.167ms of backlight on time. So with a strobe duty of 001, the backlight is flashed on for 0.167ms and remains off for 16.53 ms (16.53ms + 0.167 = 16.7).
So, strobe phase changes the timing of the strobe by 0.167ms per phase point, and strobe DUTY increases the persistence by 0.167ms per duty point, at 60hz refresh rate.
At 100hz refresh rate, WITHOUT A VT Tweak, instead of 16.7ms it's 10.0 ms.
So strobe phase changes the strobe timing by 0.1ms (since there are 100 values of strobephase, 10 / 100 is 0.1 right?), and the strobe duty persistence is 0.1ms per point.
For 120hz it's 8.3ms and 0.083 and 144hz its 6.9ms and 0.069.
As you can clearly see, the BASE persistence and strobephase timing positions (in ms) are based on the raw refresh rate times.
Using a strobe phase of 100 WITHOUT a VT tweak starts the strobe timing at the very top of the screen instead of at the very bottom (starting at the very bottom is basically a 1 frame penalty, as the strobe extends into the next following frame)--remember about refreshes being done top to bottom, right? So--lower input lag. But the price you pay is, a large amount of crosstalk (crosstalk=pixel data from the NEXT FRAME) appearing at the bottom of the screen.
Now what do VT tweaks do?
VT tweaks increase the vertical blanking interval. This is basically very similar to increasing the amount of vertical pixels (height) the screen scaler has to work with, as far as strobing is concerned. By increasing the blanking interval, you basically give the panel pixels more time to "settle" before the next frame gets sent to the monitor. This reduces crosstalk. Vertical Total 1350 increases vertical blanking by about 13% and vertical total 1500 by about 25%.
However, the scaler sees an increase in the vertical blanking period as an increase in vertical resolution itself. Why? Because higher resolutions (like 2560x1440) use a higher vertical blanking by design (this is how resolutions work--the higher the res, the higher the Vblank). That's why if you unlock the FACTORY (not service menu)--read the sticky on how to unlock the factory menu-- you'll see a bizarre resolution listed. E.g. 1920x1080 with a 1500 VT applied will show up as 1280x1440. Ignore the horizontal and look at the vertical. You'll see the increase in VT from 1138 (or 1144 at 120hz) to 1500 is the same percentage increase as 1080 to 1440.
Ok but why is this even important?
Because the problem is, the monitor's strobe timing is based on default established refresh rates and blanking intervals (1138 at 100hz, 1144 at 120hz). When you change the VT to a value the monitor doesn't recognize as belonging to a refresh rate, the monitor falls back to the "60hz" persistence and strobe timing values---0.167. So at 120hz for example, instead of using 0.083ms and 8.3ms total, it uses 0.167ms. This changes the strobe DUTY base persistence values to 0.167ms, so if you wanted 1ms persistence, you would use a Strobe duty of 006 now instead of 012 (006 x 0.167=1ms, instead of 012 x 0.083ms = 1ms).
However this does bizarre stuff to the strobe phase.
At 120hz, the entire max strobe phase timing the monitor can use for strobe positioning timing is 8.3ms since that's the frame time, right?
however the monitor is using 16.7ms persistence values! So a strobe phase of 000 ->100 would use 0.167ms x 100 ranges, but as you can clearly see, there is only 8.3ms available at 120hz. So--going past strobe duty 049 (0.167ms x 50 =8.3 !!!!!!), makes the BACKLIGHT SHUT OFF, because you are strobing past the available frame. This also limits the maximum strobe duty you can use (in reverse). Don't worry, the entire screen is still being strobed from very top to very bottom, but its being strobed at twice the width.
So at 120hz, 50 positions, at twice the width (0.167 instead of 0.083 per point).
Or at 100hz (VT): 60 positions at like 40% of the width (0.167 vs 0.1ms).
I hope you can follow that.