Buff/Debuff Overwrite Suggestion

[Shurrikhan]

That sounds a little too easy and 'hand-holding'. There'd basically be no penalty to improper timing.

One thing they could do, is allow buffs like that to form stacks, but have successive stacks have diminishing returns.
1 stack of Trick attack gives a 10% debuff.
2 stacks of Trick attack give a 15% debuff.

This way, it certainly wouldn't be worth having two Ninja's, but if you did, and their buffs overlapped, it wouldn't be wasted, but there's still incentive to time them apart.

I'd still rather not have to deal with 3+ TA icons simultaneously present on a 24-man boss. And if we'd be doing this all just for NIN, well... TA is already absurdly powerful in 24-mans and easy to time in 8-mans on the off-chance you get two of them, so it'd be far from necessary.

Personally, I'd rather take the typical approach: allow up, say, a tenth of any buff's duration to roll over from the replaced application to the new one. It fundamentally mitigate the issue of stacking (de)buffs, but it does make it easier for players to skillfully circumvent the issue.

Simply put, it allows for a small window of leniency. Replacing Trick Attack in its last second would still render its full effective duration, as would replacing Demolish in its last 1.8 seconds (about a SkS-heavy GL4 GCD), Higanbana in its last 6 (allowing SAMs to alternate between early Higan and late Higan so that Tsubame-Gaeshi isn't such a pain in the ***), etc., etc.

Pipedream:

If I had to pick a system for combined stacking, however, I'd prefer to see applications both enhance potency and extend duration, with the least flat gains to whichever dimension (rate or duration) is larger. In this way, lower potency (de)buffs aren't necessarily made pointlessly long and stronger potency (de)buffs aren't necessarily made stupidly bursty. Moreover, the game doesn't have to recognize multiple stacks as to let them fall off individually. Instead, the duration and potency would simply calculated upon reapplication, never after. Gameplay-wise, this would mean one could choose between burst and sustain almost evenly, but with burst opportunities softened a bit.

Let's take some TA examples, starting with their effects being applied simultaneously.

Single TA: 10% effect for 10 seconds' duration => .1 effect * 10 seconds = 1 effect-second.
Two TAs: 14.14% effect for 14.14 seconds' duration => .1414 effect * 14.14 seconds = 2 effect-seconds.
Three TAs: 17.32% effect for 17.32 seconds => .1732 effect * 17.32 seconds = 3 effect-seconds.

The pattern here should be obvious: 1.414 is the square root of 2, 1.732 is the square root of 3, etc., etc.

Stacked perfectly atop one another, each evenly increases the potency of the prior stack(s) and extends their duration by the difference of sqrt(n) and sqrt(n-1).

But, let's try this with stacks not granted simultaneously. I have the math for this worked out, but my explanations could use a bit more rounding out, so let's just follow the numbers for now...

Here we will have 3 TAs, each launched 5 seconds after each other. T=first Trick Attack.

T=0: First Trick Attack.
T+0 to T+5 = 10% effect --> .5 ES.
T+5: Second Trick Attack. Adds [sqrt(2)-sqrt(1)] potency and extends its duration by 5.6 seconds. At this rate, TA will last a total of 15.6 seconds, up from the normal 14.14 in exchange for having missed the .414 modifier across the first 5 seconds, at 14.14% effect, to a total of 1.5 ES (the .5 remainder + 1 for the next stack).
T+5 to T+10 = 14.14% effect --> .707 ES.
T+10: Third Trick Attack. Adds [sqrt(3)-sqrt(2)] potency and extends its duration by 5.35 seconds. At this rate, TA will last a total of 20.95 seconds, up from the normal 17.73 in exchange for having missed the .317 modifier across the first 10 seconds, at 17.32% effect, to a total of 2.5 ES (the .793 remainder on S1/2 + .707 for its effect on stack 2 specifically + 1 for the final stack itself).
T+10 to T+20.9: 17.32% effect --> 1.793 ES.

Total: still 3 ES.

So, you can burst at 17.32% from the start for 17.32 seconds or stack it up gradually (a second and third each 5 seconds later) for a total of 20.9 seconds. No matter what, you get the same nominal Effect-Second value in total -- that is to say, exactly the same nominal contribution as if you used 3 TAs with zero overlap. Just decide how the burst should best be timed. You don't get a ridiculous 30% burst, but neither do you get a lumbering 30-second duration. Gameplay-wise, it's simple and clean.

It sounds complicated if you're looking at these decimal values as if they were random, but when you just put it in the square root terms of 1, 2, 3, etc., it's really quite simple. It's basically just finding equivalent (additional) area as you shorten one dimension or the other and then allocating that across its dimensions accordingly.