While I don't think hitting 1-2-3-4-5-6-7-8-9-6-5 on Dragoon is any more entertaining than hitting 1-2-2-2-2-2-3-3-3-3-3 or really even 1-1-1-1-1-1-1-1-1-1-1, I'd actually much rather see combos made more flexible than that they be consolidated, especially without the tiered structure useful to that consolidation.

The ideal structure for key-shift play is something with an equal number of choices at each stage in decision-making (thus creating a geometric series, over n steps until cycled, of its total number of choices -- e.g. 3, 6, 9 or 3n (n=3) in the case of Monk). At present, Monk is the only job that would lend itself well to key-shift play, and even then only without considering Perfect Balance, which would have to be completely reworked.

In Opo-opo, Monk has three choices. In Raptor, three choices (True replaces Boot, Twin replaces Dragon, and One-Ilm replaces Destroyer). In Coerul, three choices (Snap replaces True, Demo replaces Twin, and Rockbreaker replaces One-Ilm).
In its first GCD, Dragoon has three choices. If he uses Heavy Thrust, he then has two real choices. Once either of those two remaining choices are used (with or without Heavy Thrust), the 2nd GCD has no choices (well, 1 and only 1). 3rd GCD - no choice. 4th GCD - no choice. 5th GCD - no choice. 6th GCD - no choice. 7th GCD - 01 to 3 choice. 8th GCD - 1-2 choices. 9th GCD - no choice. 10th GCD - no choice. Etc. Though it certainly makes no sense in its present setup, it makes little sense in key-swap consolidation either. The table of skills is simply very poorly designed. 3x1x1x1x1x1x2x1x1x1x1x1 does not a cohesive system make.

And, again, though the worst offender, it's not the only job unfit for a key-shift system. The standard Spinning Edge, Fast Blade, Heavy Swing, Hakaze, Hard Slash, etc. opening melee skills kidney shot that system from the beginning. Unless one is willing to sacrifice access to AoE skills and ranged skills mid-combo, the disjointed expansion of skill-tables following the opening ST melee (e.g Hakaze into Yuki/Shifu/Jinpu for a 1x3x1 table) prevents any fewer than the maximum spread's keys from being used at all times.

Now, if we're willing to give up that access, locking ourselves into a choice set for the full 10 seconds of combo or whatnot, then we could have something like this:
Hakaze ----- Jinpu -------- Gekko
Enpi -------- Shifu --------- Enpi
Fuga ------- Yukikaze ----- Fuga
But, frankly, even that looks awful.

A much better option, imo, would be to give each ability actual weight. Still allow for rotations, but there should be inherent value to each skill.

Make use of TP. It's every bit as much a matter of bloat at present as Dragoon's skill count. It's a resource dedicated solely to ensuring that everyone takes and presses an additional button's worth of bloat every two minuets. (Apart from ensuring that physical classes are always inferior in sustained AoE compared to casters).

Use TP as the guiding parameter for combos. Every skill can be used on its own for additional power based on current TP. You regenerate TP over time. Before adding in additional factors, the simple idea is that the more potential damage there is in a skill, the more TP it would consume. If you want to open into combat with Heavy Thrust into an immediate Chaos Thrust spread across three targets, that'd just barely be possible, but it'd cost and maybe even diminish the last of those three Chaos Thrusts. If you attempt to use a skill without sufficient TP, it cuts down its duration and potency proportionately.

Let's say we regen 50 TP per GCD. Let's use the example of a Dragoon for this. And let's say we have a TP maximum of 500 by level 50 and beyond. Here are your TP costs:
True Thrust - 0
Impulse Drive - 20
Vorpal Thrust - 100
Disembowel - 120
Full Thrust - 240
Chaos Thrust - 270
Each action reduces the cost of skills deeper within its combo set 50%, stacking multiplicatively.
Thus in a True Thrust combo one would generate 50, 0, -10 TP, for a total of 90 TP regenerated over 3 GCDs. That means you could afford a free Full Thrust every 2.66 FT combos, and a free Chaos Thrust every 3.33 FT combos.
An Impulse Drive combo, likewise, would generate 30, -10, -20 TP, for a total of 0 TP regenerated over its GCDs. Thus, if rotating CT, FT, FT, FT, would regenerate 270 TP every 4 combos, enough for a free Chaos Thrust every normal use of Chaos Thrust.
Note that if Heavy Thrust weren't a thing, CT-FT-FT-FT;repeat would perfectly time to Chaos Thrust's DoT's replacement at a 2.5 GCD. 30 seconds / 2.5 seconds per GCD = 12 GCDs; 12 GCDs / 3 steps per combo = 4 combos (per 30 seconds).
So, every time you'd normally be able again to Chaos Thrust you could Chaos Thrust an extra time. And every 3 rotational strings at level 50 by which you'd refresh Chaos Thrust, you could (if maximum TP were high enough) replace the whole of an FT combo (TT-VT-FT) with 3 back-to-back Full Thrusts.

There, now you have variety available to rotation, a bit of added flexibility, a way to manage your own TP to bring melee ramp-up more in line with that of casters, and no need for a two-minute cooldown. And with all that you can give each skill importance of its own, e.g. Vorpal Thrust for pushing enemies to the side via paired WASD / AnalogStick movement, Full Thrust for knocking enemies into the air, Disembowel for its auto-parry component and its bonus counterattack damage, etc., etc. Now you finally have a functional and meaningful use for all of those weaponskills and for the TP system restricting them.