The suggestion's been thrown about before, but just keep in mind that unless the Shield GCD is especially long, it will just end up working like an oGCD, and if it is especially long, it will end up forcing you to choose between using Shield Swipe (assuming it's moved to the shield GCD) and Shield Bash readiness. Should its animation locks be extended, then it will just end up feeling like an Empyreal Arrow you cannot optimize. Alternatively, let's say they have a longer animation lock only when followed by Sword Moves (can rapidly shift from a Swipe to a Bash, but there's a bit more delay in going to a Fast Blade, etc., similarly to if a full GCD had been used). At that point what you have isn't so much a second GCD system as simply a new way of performing combos, wherein moves within the combo recover from the GCD faster than those outside the combo, allowing for faster sequential action as part of the combo benefit.
The first issue is that you'd need at least a decent variety of Shield skills. For instance, if Shelltron were moved to a no-CD forced block channel that absorbs up to X damage via blocking and uses some alternate charging resource and renamed simply to Guard, then you'd actually have some decisions to be made between pacifying via Swipe, stunning via Bash, and blocking via Guard. Minor added complexity.
The second is that these Shield choices would have to be relatively frequent or regularly accessible choices. This could be done by removing the on-block requirement from Swipe, instead improving it or refreshing its cooldown under that condition, and such things as the above "Guard" (Shelltron replacement) example.
Third would be that those choices would need to be viable, but tactically distinct from sword choices. For instance, let's say you take out the rigid combos. Rather, any Sword or Shield combo initiator or bridge generate benefits for either, to be consumed on the next in line. Neither cancels combos for the other (except maybe Shield Bash). Alternating the two allows for greater build-up, while using one side back-to-back allows for more immediacy of effect. Swipe now absorbs a portion of enemy damage, doubled if against you or whatnot as not to be too powerful of OT support, for instance, and Bash can potentially be used as a nuke hit, while Riot Blade or Bash can be turned into a knock-back, and certain abilities can cause an AoE Swipe, etc.
And then there's the issue of resources. Due to the second GCD system, a PLD would have a potentially absurd rate of fire, depending on tuning. Resources would likely have to be spent to compensate, limiting longer term potential, but that would still leave it bursty only to fall into lethargic TP starvation. And this is the tank that already most suffers from TP issues. Alternatively, you'd have to develop a new resource entirely that doesn't act as a simple T-3 minutes to 'inevitable disappointment'. Skill Speed adjustments to remove oversights, similar to the recent Sword Oath change, could certainly help, but I'd imagine the second system would be necessary as well.
All in all, if you want a more interesting style of play, this would be only one of many ways to go about it. You could always adjust the regular rotations, swap out certain skill effects, and so forth for improvement in tactical feel, urgency, skill-cap, etc.