Resistive loads and contactor vs relay debates... I find speaking to many sparks and automation engineers, that the sparks lean towards contactors, yet the automation engineers seem happy for relatively low loads to use relays.
I suspect this is because modern relays are very capable and in the field prove remarkably reliable. They probably do give up before contactors but the benefits of size, quietness and low cost probably more than compensate. And of course, a contactor is just another type (category perhaps) of relay. They were developed when contact materials were less advanced and understood and it was a simple solution to simply have a larger, stronger contact area and breaking spring than in an equivalent standard relay of the time. But who's to say if some modern relays can't out perform a 20 year old contactor design now?
We use relays to control special FX work for live TV we sometimes put two NO relays in parallel, so that should one fail we still get an energised circuit on cue. We back that up with a kill switch in case we need to de-energise the circuit in the event that a relay sticks. We do this for obvious reasons, we have to know that when we hit the go switch, things will happen. But the truth is, over literally thousands of activations not one relay has failed yet. We smashed one to pieces once to see what was happening inside and you could see some darkening around the contacts, but it was very slight.
I should add that we do use flyback diode's, so the relay is fairly well protected during it's duty cycle.
Just offering the above as input based on what we have experienced. Perhaps one solution would be to split the cold tails from the UFH into 2 circuits, and place each on a suitable rated relay - then switch both those relays using a third, itself switched from the stat? Or switched directly from the stat if more convenient. This would also avoid a complete heating failure if one relay did give up. I feel that it's not necessary to use a large, noisy contractor when you're actually switching multiple UFH circuits (as in laid circuits of heat mats) at once anyway, why not just split the load into separate relays and enjoy a quiet life with built in redundancy?