One dark output tube on 100w Marshall... Haunted Socket?

[Packerswin14]

I have an odd situation that apparently has happened to other people, I've seen it mentioned elsewhere on the internet with no solutions.  I haven't found any posts here, but maybe I just haven't come up with the right parameters to search.  I'm working on a Marshall 100w head that some guy modded, but it's still a basic 100w Marshall circuit, 4 EL-34's.  The 4th EL-34 closest to the preamps won't function.  The filament won't heat up.  I've checked continuity in the heater circuit, and there's continuity all the way through the circuit, from the PT, through all 4 output tubes, and through to the preamp tubes.  I've measured the voltage with the amp on and no tubes in, and there's 6.3 VAC all the way down the line.  I've replaced the tube socket.  I replaced the pieces of wire, thinking somehow there could be an internal break, but that doesn't make sense:  The 6.3V is getting beyond that tube to the preamp tubes, so why isn't this one tube functioning?  I've tested the tubes on two known good testers, a Mercury 1000 which does a good job for shorts and leakage, and an Eico 666.  I swapped the tubes around to different spots, and then tried two other sets of tubes of two different brands.  I've reheated solder joints... With the two outer EL34's removed, the amp functions ok.  It has functioned, on occasion, with all 4 tubes in, but then there's added annoying hum, but mostly that one spot won't heat the filament.

I've seen mention on the 'net that there's a mysterious problem with this circuit, anyone ever seen such a mystery, or am I just missing something obvious?   Any help would be much appreciated.

[The_Gaz]

That's beyond strange. You changed the socket, tube, and wiring. What else more could you do? At this point the only thing I can suggest personally, is to wiring the power tubes in a different order, just for the hell of it. You also might as well try another socket and tube!

[sluckey]

Measure the voltage between pins 2 and 7 of the suspect socket from the 'tube' side of the chassis. The filament string is connected to pins 2 and 7 on the bad socket, right?

[Packerswin14]

Oddly, there is a 6.3v AC reading at the top of the socket, too, I was sure that was going to be it.  The guy I was working on it for is going to send it back to the guy who modded it for him and make it be his responsibility.  I may never know the answer to the mystery.  Thanks for the help!