My concepts are kind of "belt and suspenders," in that an expendable layer of ablative figures in the design to back up the water cooling in case of failure. More weight, but there it is. While an ablative is difficult to replace and maintain on a reusable vehicle, it is acceptable as one-time emergency backup.

While redundancy is usually a good thing, sometimes it can be counterproductive. Back when Bill Lear's engineers were designing the first Learjet, he told them to not put anything into the plane that wasn't absolutely necessary. His words were, "If you don't put it in there, then not only do you save the weight of it, you save the weight and expense of the backup."

Lear's directions were a follow-on to Geoffery deHavilland's dictate to "Simplificate and build in lightness." If the backup heatshield is strong enough to be an effective backup, then its strong enough to be the primary (or sole) shield. In that case, there's no real need for the water cooling. Leaving it out will save weight, complexity, and expense.

If a capsule design is chosen (and I thing there's good reason to do so), it should be possible to build a capsule with a replaceable heat shield if the capsule is intended to be reused.

Like I wrote before, there are multiple approaches to consider for a new manned spacecraft.

1. Fully reusable - this means that must be able to be refurbished quickly and safely for the next mission. The goal is to lower per mission costs over the long term at the expense of higher development costs. Done properly, a fully reusable system can end up costing less over the long term (pity that hasn't been the case with the Shuttle).

2. Partially reusable - this could be done with a capsule with easily removable systems like guidance, avionics, etc. The capsule shell might be tossed after a mission and the really expensive systems get flown again on a new capsule.

3. Disposable - this is the model used by Mercury, Gemini, Apollo, and probably Soyuz. Systems only get flown once, so there isn't the higher R&D costs of creating reusable systems. However, this means that there's higher requirements for system checkout and test before each mission because every component is new and therefore unproven.

The two driving factors should, IMO, be overall safety and life-cycle costs. While in theory a fully reusable system should win on both of those counts, it might work out better to go with option 2 above.