Here's the thing: Customization (especially the likes of "Talent systems" outside of few exceptions like Rift) has historically almost always constrained gameplay options --replacing that gameplay with menu-play-- rather than expanding them. And I don't just mean this in a sense of "all choice is illusion" or similar BS; we don't all just play the 'best' job per sub-role.
Consider it this way. A game has n undermechanics that it must support player interaction with. Now, if there's limited button-counts and button-efficiency (e.g., Stormblood/Shadowbringers level button-bloat, where virtually none of the obvious opportunities for consolidation were taken), that can be bad, because each job's kit has to be able to engage with those undermechanics, even if with varying ease or directness, and that may leave less room for depth*.
*(Or, in XIV's case, sets of 3+ buttons to use all together or in rigid sequence as basically one action, which is more akin to rhythm-gaming than actual depth, but still.)
So, in this pre-specialized design scheme, every job, regardless of role, would want to be able (from within virtually any composition, without any prior role-allocation) to take some part in...
- manipulating enemy behavior,
- intercepting/thwarting attacks for allies at greater risk,
- crowd control,
- Stagger,
- suppression,
- timely defensives,
- Elemental systems,
- coordinating bursts of damage,
- knowing when at-cost uptime (standing in fire) is worth that cost,
- manipulating the costs of engagement,
- providing the resources to enlarge the range of viable options,
- etc., etc.
And while that may leave less room for your separate Senei and Guren, Shoha and Shoha II, Ikishoten and Namikiri buttons, etc... the job would tend to therefore have more that it can do and play with.
Now, let's specialize those jobs. A "Tank" becomes so much better than the rest at manipulating enemy behavior and defensives that everyone else's capacity for such becomes redundant as is trimmed shortly thereafter, regardless of their potentially serving a double-purpose; at best, only tiny little rhythmic bonuses remain (via Third Eye, Feint, etc.). A "Healer" makes so much better use of restoration that intercepting/thwarting attacks for more at risk allies, providing others with the resource for sustained or broader action, etc., likewise becomes, in majority, redundant.
The various gameplay elements that previously made a party-wide mechanic of reducing damage taken while maximizing output ("tanking") then becomes replaced by "a Tank exists". Matters of sustain and risk-calculation (with consequence) are largely replaced by "a Healer exists" (and simply the cost to their offensive ppgcd, since there will always be resource enough to heal you).
In pruning those now redundant 'versatility' tools, arguably there'd be room for more complex layering of one's rotation... but I have yet to see that ever be the result of pruning.
Now, that specialization would usually be called Roles, ofc, but that is essentially the endgame of customization: to be the best at something, which in turn tends to make everyone less "best"... redundant, essentially snapping any at-cost options towards very particular norms that section off gameplay.
That customization typically means no more of everyone having access to A, B, and C to any degree that'd promote coordination (as opposed to each working individually, but in proximity to each other -- without the weight of, say, this healing then being more or less worth on any basis but whether otherwise fatal damage would immediately follow, etc., etc.). Instead, A goes to Tanks, B goes to Healers, and C goes only to Damage-dealers.
Similarly, whenever Talent-based or similar customization allows for remotely excessive specialization, you end up with shit like... Single-target Execute damage goes to the Assassin builds of Ninja, forcing them for this particular fight A and out of B and C; Super-Flare build takes add fight B, but is barred from A and C; etc., etc.
Customization has, in short, a pretty fine edge between even slightly increasing net gameplay choice and significantly decreasing it, especially when concerned at all with what may be made obligatory (or, how many of those "choices" will, in practice, be non-choices). Part of why its so expensive to develop is exactly that: unless limited from the start to largely cosmetic changes, it can slip quickly from any sort of benefit to a net loss, which is not normally something desirable for a system that itself would come at significant development cost.
And yet, despite all that, I would like some further customization in this game. It's just very worth knowing how and why it can go wrong, often to the point of (relative to other models of gameplay expansion) letting players choose what little they want to keep, instead of choosing what they'd like to add.
That said, I don't think a Talent system (unless you're imagining it rather differently from the likes of WoW, GW2, Rift, etc.) would be a remotely efficient investment. I would suspect instead that a "All Jobs on One Character" game ought to leverage that for its customization, rather than creating a ton of sub-jobs (via those different builds) in isolation from each other.