The 'problem' with TVSs is that they are not perfect clamping devices.
Firstly they have a tolerance that you can do nothing to change (other than selecting another sample). They also have three important parameters:
1. The holdoff voltage. That's the voltage where none of the devices, over the full manufacturing tolerance spread will conduct any significant current. This is the maximum voltage of the rail you would use them on.
2. The conduction voltage. That's the voltage where the TVS will conduct a specified current ( just like Vz @ Iz for a zener diode).
3. The clamping voltage. This is the point will conduct hard, again at a specified (large) current. This is the point where your fuse would start to blow.
In practice, for a 5V (holdoff) device, the spread of these three voltages will be quite high, you probably won't hit the clampiing voltage until about 8.5V... probably higher.
In contrast, your crowbar circuit with its nice accurate TL431 and trimmable resistive divide can be made to trip at, say, 5.5V. Even if you replaced the Thyristor with a big beefy transistor to act as a voltage clamp rather than a crowbar, it would still perform better than a TVS in your situation.