Then tell me what is at the junction of C7, C8 and T2.
While Dons last explanation winds a RF ground path thru C5,6, and 8, 'splain to me how that even comes close to good engineering by placing some level of the plate RF on the screen pins?
... And why in hell would anyone use an RFC to the L5 CT and then feed RF in to the bottom end that has to wander around before finding some sort of compromise ground?
All I see at the junction of C7, C8 and T2 is the rf choke L4. But L4 doesn't place any plate rf on the screen pins. The other end of the rf choke is connected to the mid-tap of the balanced tank circuit, which is at precisely a cold spot on the coil, rf-wise, so the bottom end of the choke doesn't feed any significant rf to ground. The purpose of L4 is to carry the DC plate current to the coil, which in turn carries it to the tube plates. Since the tank circuit may not be perfectly balanced, there may be a slight amount of rf voltage at the mid-tap, but the series inductive reactance imposed by L4 and the by-passing effect of C5 and C6 effectively remove any trace of plate rf from the screen pins. The second function of L4 is to isolate the mid-tap of the coil where it is connected to the +HV, from the rotor/frame of the split-stator capacitor. You
never want to directly connect the midtap of the coil to the rotor of the capacitor, hence the rf choke.
There is undoubtedly more plate RF on the screen pins via the internal screen-plate capacitance in the tube than from the midtap of the coil, through L4 and C8, on the way to ground through C5 and C6.
I also dont buy into the wandering explanation about grounding and an absolute need for a common point ground at low ham frequencies.
Did you ever try to get the parasitics out of a 160m-20m push-pull, cross-neutralised, triode final?
Heck, that circuit doesnt even have any parasitic suppressors which limits its frequency anyway to low bands where the 813 parasitic paths are effectively swamped.
I do find that questionable, and I wonder what year that write-up was done. I have
never been able to build a final for the low ham frequencies without parasitic suppressors that was not squirrelly, whether using beam tetrodes or triodes, link-coupled or pi-network. Art Collins introduced the concept of parasitic suppressors some time about 1934 with one of his production amateur transmitters. What was so amazing about that rig was how smoothly it tuned up. The secret was the suppressors, which consisted of the familiar composition resistors with a few turns of wire wound round them. Up to that point, the suppressors were absent from ham construction (look over the circuits described in the old Handbooks). Hams just assumed that transmitters were inherently squirrelly to tune up, and put up with the annoyance, doing the best they could with what they had, until the Collins rig proved otherwise.
I also detest an unswamped by a carbon resistor RFC in the grid circuit and always do that when substantial grid current is involved and a 2-3K carbon resistor alone is impractical. With a RFC and a HF effective value bypass the TGTP possibility is slim to none but it still opens the possibility of a LF oscillation which can often go undetected without an SA scan.
The TGTP oscillation caused by an rf choke in both the grid and plate curcuit almost always is LF, or near the operating frequency, not VHF. With a push-pull amplifier like the one here, there should be no significant rf grid current through the grid resistor, just DC, since the grid resistor is connected to the mid-point of the balanced grid tank which should be an rf cold spot. With a single-ended grid tank, the cold end of the coil should be by-passed directly to the common ground point with a capacitor, and another by-pass capacitor should be added to the other side of the grid resistor as well, where the lead to the the bias supply and metering circuits is connected. There is no reason to use an rf choke here in any case.