I thought of yet another possibility: Drill a hole in the manifold and weld on an O2 sensor bung (or use an existing socket). Make a heat probe, either machined from solid copper or by using an old O2 sensor body, that puts the end of the copper heat pipe directly in the exhaust flow.
Make the receiving-end of the heat junction out of a heat-conductive metal with a higher rate of thermal expansion than the heat-sending end, so when it gets hot (coolant at operating temp), it expands and breaks (or at least reduces) the heat transfer path. Copper plug (thermal conductivity 400) on the sending end (coefficient of thermal expansion 9), aluminum socket (thermal conductivity 250) on the receiving end (coefficient of thermal expansion 12). Machined aluminum block that has a socket for the heat pipe and a coolant path. Maybe some heat-transfer fins in the coolant path to improve transfer, maybe not. Spring clip to hold it all in place.
When it's all cold, they may or may not be clamped together-I'd have to run the numbers. As the copper warms and expands, it squeezes against the aluminum, which transfers heat to the coolant. As the coolant reaches operating temperature, the aluminum warms and expands enough to relax the contact with the copper, slowing heat transfer when it's not needed.
So there you go, that's the most elegant solution I can come up with. No break in the coolant path, no delicate parts, no moving parts. Failure mode would either be regular old hose clamp leak, or the copper plug breaking off. First is easy to fix, second doesn't have any ill effect beyond stopping the heat transfer. Installs with a 7/8" O2 sensor wrench and a screwdriver. Pointing out galvanic corrosion issues at this point would just make you look like a spoilsport.
