Thank you all. I don’t believe it is a burned valve. The plot lacks the periodicity that Bush shows. Besides, it would not just disappear. Below is a zoomed in plot.
Stuck valve is one of my fears (yes IO540), and I try my best to mitigate it by regularly changing oil and flying at max EGT to a tiny bit on the lean side to avoid coking, which feeds build up on the valve stem. Reading up on various forums the Lycoming SB is seen as questionable. I signed up with SavvyAnalysis, and am now in contact with them. Will report back my findings. I think its a pretty cool service, sitting on a treasure trove of engine data, perfect for machine learning algorithm to party on.
If you see EGT well over 1600 at say 65% power ie 23 2300 12gph that is physically impossible on 100LL and 8.5:1, with or without exhaust valves 
Maybe pre ignition?
I see 1600 when doing lean mag check at low power, say less than 55% . I think it is possible if combustion is retarded so much that the flame front escapes the exhaust valve and hits the probe. In the plot, where EGT exceeds 1600, fuel flow is 12.1 gals/hr, so about 70% power. Let’s say valve is stuck open, what would the combustion actually look like?
On half mag, you get a higher EGT, I believe not because the gas is hotter but because the mixture is burning slower so you get the still burning mixture reaching the EGT sensor which is some 10cm down the exhaust pipe.
JP Instruments says “nominal distance” 2-4 inches from the exhaust flange i.e. 5-10cm. The important thing is to make sure the distance is uniform.
As the previous advice has been given, the thermocouple probe sheaths get eaten (corroded) by the exhaust stream and wear out over time. Before getting worried about this and getting caught up in considering stuck valves and data analysis, just switch one of the channels and see if the erratic one moves to the new channel.
I use thermocouples a lot with data acquisition at work and anytime we do a set-up with multiple probes there is invariably a duff one (mind you we are typically quite hard on the set-ups so there is a high attrition rate). Consequently, it is not surprising to see an erratic channel.
Also, no harm in buying a spare thermocouple or two, you are going to need them eventually. If one is starting to go, then the others might be on their way too! I seem to recall replacing some in a Cherokee I owned and think one or two of the replacements was defective (straight from the manufacturer-so much for quality control!).
A search for e.g.
defective jpi probe
digs out various relevant threads.
I too had a duff probe from JPI – brand new. I sent it back to them and they washed their hands of it, saying they could not tell where it came from etc. Well, it’s true the probes don’t have serial numbers.
