The guys who teach the Advanced Pilot Seminars in Ada, OK have an extremely sophisticated engine test lab (probably the most advanced in the world for aircraft engines) where they can do all sorts of tests. They originated the recommendation to try to keep CHTs below 380* and Mike Busch has helped to spread that word far and wide. The rationale behind it is interesting, and somewhat obfuscated by Mike's talking points.
Yes, there is a significant decrease in the strength of the metal alloys in the cylinders and heads as temperatures go up. The "rough" numbers (rounded slightly for ease of discussion) are that at about 400*F, the alloy has lost a bit over 1/3 of its "yield point tensile strength" measured in PSI. At about 450*F, where Lycoming typically sets the bottom of the "yellow" band, the alloy has lost over 1/2 its strength. At 500*F, it is down to less than 1/4 of the original strength. No one I've ever heard of has any argument with that data - it's generally accepted by metalurgists for many years.
Where the argument begins is at what point do we enter the "danger zone" for engine operations? Several factors influence that analysis, and one of them is absolutely HUGE.
If you're using a spark-plug "ring" thermocouple to measure CHTs, the APS folks have proven that those probes read easily 20-50*F HIGHER than a bayonet-type thermocouple inserted into the CHT measurement ports ("wells") in an aircraft engine. It seems that most of the "factory" supplied gauges in the past (when the vast majority of airplanes were certified and built) came equipped with a single-point CHT (1 cylinder) that used the ring-type thermocouple installed under the spark plug of the "usually hottest" cylinder. Since those ring-type thermocouples typically indicate higher CHTs than those measured by the "well-type" thermocouples (generally accepted as being more accurate), both Lycoming and Continental (and by extension Cessna, Piper, Beech, etc.) appear to have made allowances for that difference in setting the "yellow-arc" and "red-line" temps for their engines. Thus we often see "book" limits of 450*F for continuous operation and 500*F for "do not exceed" limits.
If you are using an all-cylinder digital engine monitor on your engine, as Mike Busch (and the APS folks) STRONGLY recommend, it is highly likely that you are using the well-type thermocouples on all the "newly equipped" cylinders. Some will retain the "factory" ring-type thermocouple on the originally monitored cylinder, but it is not recommended by JPI, EI, Mike B, the APS guys, or pretty much anyone else. Use the same "well-type" probes in all your cylinders if you're spending the money to be monitor them. Why would you want one cylinder probe that constantly reports a known higher temp than the others? So, moving forward with the assumption that you now have a more accurate indication of the actual temps, when your digitally-monitored engine reports CHTs of 450*F, it is highly likely that the cylinder head temps are ACTUALLY around 450*F. When you were using the factory ring-type temp probe, a report of 450*F would almost certainly mean that you were actually running somewhat lower temps - closer to 400*F.
Look back at the paragraph about the effect of temperature on alloy strength, and note the difference between 400*F and 450*F. Pretty dramatic, right? Note also the difference between 450*F and 500*F. The APS guys determined that between 450-500*F, cylinders became more "plastic" and began to go "out of round" resulting in scuffing of the cylinder walls. Once temperatures actually rose above 450*F, the CHTs could rise so quickly that even dramatic pilot intervention could not bring the temperatures down before CHTs went WAY above the 500*F point. And going above 500*F for any length of time can be absolutely fatal to the cylinder. They hypothesized that the "mental" red-line (as measured with well-type thermocouples typically used with digital engine monitors) should be adjusted to 450*F to prevent those issues, and likewise the mental "yellow-arc" should begin at about 400*F.
To validate their hypothesis, the APS folks began doing a whole series of tests (mostly with Continental engines, but validated with Lycoming engines as well) that showed that for most pilots, when "high CHT" alarms were set to alert the pilot as CHTs went above 400*F (under conditions that might lead to continued rise of those CHTs), most pilots could respond in sufficient time to prevent excursions above 450*F, but only barely, and only if they were focused on that task to the exclusion of other tasks. Add in "normal pilot workload" that you typically see in high-power situations (takeoff, initial climb out, go-around, etc.) and the pilot's reaction was slowed enough that they might NOT be able to catch the rising CHTs in time to prevent exceeding 450*F. Thus the APS folks recommend setting the initial "high CHT" alarm on your digital engine monitor to alert the pilot when the CHT increases above 380*F, and that pilots take immediate action to reduce the CHTs (lowering the nose slightly in climb, etc.) and to closely monitor CHTs once that 380*F number is exceeded. If the CHTs continue to rise towards the 400*F number, they strongly recommend the pilot be prepared to take whaever action it takes to correct the problem. That typically means significantly enriching the engine (perhaps even going to full rich at lower altitudes), but may (in the case of LOP operations) mean leaning even further (once you go LOP, leaning the mixture further results in reduced EGTs and CHTs).
Mike Busch attended the APS, but in his recommendations he often glosses over the part about "if you have an all-cylinder digital engine monitor" when he makes the recommendation that the APS also makes - keep the CHTs below 380*F if at all possible. The APS guys would observe that if you have a factory, ring-style CHT probe, it is highly unlikely that you'll be able to keep CHTs below 380*F in climb or other high-power situations, since that would mean the CHTs might be actually somewhere closer to 330*F, which is simply unrealistic. Mike's advice is GOLDEN - but only if you have the well-type CHT probes. If you don't, it would seem that the "book" limits (perhaps reduced by 20*F to allow for "pilot response time") would be a more realistic goal.
If you get the chance, the APS seminar (
http://www.advancedpilot.com) is probably the absolute best value for your aviation dollar. I thought I was reasonably knowledgeable about engine operations before I took the course, but I learned so much it was incredible. There is an online course and a "live" classroom course. Online is less expensive, and after you take that course, they will credit you the full cost towards the "live" course. I did both, and although I learned a LOT from the online course, the live course was FAR more valuable. It was, quite simply, the best-structured educational experience I've ever had - and I spent an entire career in IT, where ongoing training and education is mandatory if you want to remain relevant. I've been to literally thousands of hours of training courses, and the APS was the best I've ever seen. Well worth the cost, and if you currently have a digital engine monitor, the APS course will probably save you more money on maintenance over the next 2-3 years that it costs you.
I know the knowledge I gained in the APS course may well have saved my life when I used what I learned to figure out that I had an otherwise undiagnosed induction system leak on my turbo-normalized fuel-injected engine. When we investigated that induction system leak, we determined that leak was actually caused by a pinhole exhaust leak (completely hidden by the engine/exhaust structure, and with no visible discoloration even viewing with a mirror and light) that had not only cut through the intake gasket, but would likely have eventually have burned through a fuel line with disastrous consequences... My mechanic and I just looked at each other for about 5 minutes, with him apologizing for missing it, and both of us wondering how he could possibly have known without the indications I saw from the engine monitor... and without the knowledge I had gained from APS to interpret that data.