How and Why I Changed It
The original Collins thermal sensor is kind of neat: It's a normally-closed thermal switch with a heater, which biases the thermal-switch up toward nearly opening... at this point, a delicate dance ensues: heat-calories from the tube try to open the switch, while heat-calories are removed by the blower-airflow.
BUT - it does not need to be such a delicate dance - the tube anode-seals require <250degC, and should be operated <225degC; in my prudence I think <200C should be safer. Infrared/laser remote temperature measurements suggest the tube may only rise xxxdegC above ambient, providing a very large margin, and a large window between "operation" and "danger". My solution will be a thermal-switch, which will open the 12V and K203 when it senses 160degC - ample safety for the tube, but not being a nagging nanny to the operator :-)
This alone will ensure tube-safety, but let's go even farther - let's use both belts, and suspenders :-) Modelling my Dayton-motored blower suggests the blower will produce somewhere around 0.8 inches water-column (wc) with the 4CX1500B. The published requirements for the 4CX1000A pressure is 0.2"wc at a full 1kW dissipation; the 4CX1500B requires even less at 0.18"wc for 1kW dissipation! And normal operation will duty-cycle / time-average the dissipation down.
So, we'll sense the air-pressure at the base of the tube. Again, we have a wide window to allow full operation, and also maintain total tube safety. 0.1"wc should be "sufficient" for normal operation; 0.3"wc should allow for "no-time-limit" 1kW dissipation (maybe my RTTY!).
Thanks to induced-draft furnaces and hot-water-heaters, the HVAC industry has a wide selection of suitable temperature- and pressure-sensors. On the other hand, the Collins sensor is UnObtainium, and mine doesn't appear to be working correctly. We can now have deterministic, sustainable, reproducible and improved tube safety!