Feasibility aside, I'd probably start every tank damn near from scratch, focusing on a couple of core elements and expanding from there as organically as I can (as to maximize both total depth and depth per button, each in as satisfyingly cohesive a manner as possible). I'd of course keep an eye on parity of capacities, but if each reaches those capacities in very different ways... all the better.
Some spitball base concepts:
- Paladin: Sword and Shield
- Dark Knight: Blood and Shadow
- Warrior: Wrath and Abandon
- Gunbreaker: Charge(s) and Ammunition
For Paladin, for instance, the idea of the "sword" might take on a few different ideas if looking across its level span in isolated bits, but it's always be juxtoposed against the idea of the "shield". The sword might be pure offense (opposite defense), or defense/playmaking through offense (versus shield's resource-generation or meaningful spenders, including what would amount to offense through defense), or about flow or action-based momentum (opposite banking and buff-based ramping). From the start, "shield" would play almost as large a part, rotationally, as sword skills, even if they might have slightly fewer buttons, but everything ought to develop with that thematic pairing (and gradual shift in boundary or slant between the two), rather than just slapping on more, generally separable skills for raw potency or mitigation. Rather than abandoning any earlier theme or core, it just expands to find more and increasingly synergetic ways to play around and optimize upon those themes.
Doubtless, a lot of those iterations would deserve to be more pervasive than single skills' effects, and would therefore better belong as traits than as mere parts of skill tooltips.
Warrior, meanwhile, would be all about building ever stronger and sturdier, always deepening into the Thrill of the Fight (with some option of exhausting oneself for burst at cost to sustain or overall damage). Just as rough illustration, imagine if Deliverance and Defiance -- or at least their varied spenders -- were returned, but such that spending gauge on either would provide a duration and potency of that stance's benefit based on gauge spent, rather than those benefits being passive or lost upon swap. Inner Beast and Steel Cyclone, for instance, might generate % lifesteal and increase maximum HP, while Fell Cleave and Decimate might instead increase your damage to affected enemies and your critical strike chance. (In this case, any effect like auto-crits would deal greater damage based on your excess crit chance.) In this way, your best response to both an upcoming tankbuster or even to maximizing damage to be dealt might instead use both stances, prepping maximum HP before skills like (Stormblood-era) Upheaval or increasing your lifesteal percentile before nuking an enemy with an auto-crit Fell Cleave, or even buffing the heck out of your crit chance before using that for a proportionately empowered auto-crit Inner Beast and its converted self-healing.
Etc., etc.