Another simple test which occurred to me after posting is to short out the fx loop with a plain jack lead. Sometimes fx sockets can get noisy as the switches get dirty. Putting a lead in removes those switches from the circuit. If it does turn out to be that then they just need a clean up with maybe a contact spray.
Assuming you did all of this with the Master Volume turned all the way down to remove the preamp as a source of noise then that's all good info.
So we know it plain cannot be a problem with the preamp valves as it remained even when all of the preamp signal was removed with the MV set to 0.
The TSC bias readings are all in the zone so that shows us nothing. Incidentally you may have a better match if you put your 2 x "9"s as a pair and the "10" and "11" as the other pair. (Now that is purely pedantic and won't be heard, just an observation.
)
Swapping the inner and outer pairs of valves and running it at 18W shows that there isn't a single valve of the 4 EL84s causing it. At 18W the inner pair alone would be producing the noise. After swapping, at 18W it is now the original outer pair still giving the noise.
Swapping the Phase Splitter valves shows that it probably isn't that valve at fault. The difference in noise levels could easily be due to gain variations in the two valves.
So everything seems to suggest a different source of the noise other than the valves, and if that is so and it isn't the fx loop sockets it would need equipment to track it down. The only other simple thing I could now suggest is to swap out the valves with others you may borrow from somewhere to be absolutely sure. There is the possibility that you have more than one noisy valve and the swaps are somewhere changing one noisy one for another. That would be really bad luck and is pretty unlikely! My advice would also be, don't just stqart by blindly swapping out every valve as the "revalving kits" suggest is wise. That's unnecessarily expensive and poor engineering. You won't understand your problem and be able to deal with it when it occurs again, perhaps in a different amp. Swap out the single inner pair of the output valves and try out only that pair at 18W. If the problem clears then one of those output valves was noisy. Stepping up to 36W with that new pair in place and using the outer pair will then test them as well. No noise with all four then tells you the second pair are fine and it was a single faulty valve, if not then replace the outer pair too. Job done cheapest way! Swapping the phase splitter is a single valve issue which may help to rule out the slim chance that both of your 12AX7s are noisy.
A bit more info which might be useful....
This is technical but please bear with me, it may well be relevant for us as GM users. Preamp valves, despite what current "common knowledge" says, are not over stressed in a guitar amp and they usually run perfectly for a lot of years barring physical damage, certainly a lot longer than the fanatic valve rollers say is good. Power amp valves do wear out, but they also usually give years of use before they suffer badly enough to warrant changing. Again against "common knowledge", they don't "wear out" by use. The idea of "cathode stripping" which is often cited as a cause doesn't apply to the class of valves we use in guitar amps, that's both preamp and power amp valves, the energy levels are way too low to cause it. They usually die by cathode poisoning creating "interface resistance".
Interface resistance happens when a valve is left powered on at the heater but with no anode voltage applied. With no current flowing a barrier layer slowly forms between the base metal former of the cathode and it's oxide coating which increases in resistance and gradually reduces the valve's effectiveness. This process is irreversible. Let's see, heaters on, anode voltage removed, errrmm, isn't that called "on standby"? Yes, in fact the usual type of standby switch circuitry actually harms the valves it says it is protecting.
Now worse than that, as GM users we have the power soak setting on our amps. What does that do? It starts by taking out of circuit one pair of the valves. And I believe we are right in saying that it leaves the heaters on with the anode voltage removed. So it would seem all the time we are running at less than 36W we are poisoning one pair of our output valves. That worries me and I am intending to ask H&K what they think of it. When not in use I now leave my amps on 36W and muted with the MV turned fully down rather than on Standby. I can see no penalty in doing that and it may be preventing damage to my output valves.
So the path onwards from here, assuming it isn't the fx loop sockets, depends on whether you can get your hands on a borrowed set of known good output valves and a good 12AX7. Swapping out is the only easy way I know of to diagnose this problem. Can anyone else come up with any other simple ideas?