Paralleling TVS isn't so bad. Up in the surge range, their ESR is a relatively large part of the voltage drop.
Small surges (low currents, modest voltages, long time periods) will experience imbalance, though given that the tempco is positive, it need not be destructive anyway. Avalanche breakdown is different from forward bias.
NTCs wouldn't seem to be useful for a suppression system. Nor PTCs; they take an extreme amount of energy to open. Use a good old fashioned fuse.
Some details about what application you're considering would be helpful. If this is for mains, TVS aren't used there; lightning strikes deliver too much energy for that to be safe or practical. If you need strong clamping in that type of environment, a multistage suppressor would be desirable: perhaps using some very large MOVs to take the brunt of it, then resistors and TVSs to attenuate the remaining overshoot, and RLC filtering to dampen the transients due to stray inductance.
If this is for automotive, common mode isn't a big concern. You can always just shunt stuff to ground, and I don't think you have to worry much about "-V" being other than GND, as long as it's bonded to the engine or chassis at some point -- though that can potentially open up as well, to be fair. The nature of surges is different: lower voltage, modest current, very long duration (e.g., load dump). Everything else is small electrical switching noise type pulses, easily filtered or clamped.
Tim