Real transmitters (tube type) are resonant , not arc welders... Thought I'd tap in with an otherwise worthless anecdote. Years ago when I knew even less than I do know I was told, "Real generators are welders" . This had to due with 100% duty cycle, maintenance schedule and reliability. This has, for decades served me well. I suppose I should get to the point. I have what used to be a Heathkit HA-10 Warrior. I say used to be because it now has a transformer to support 4 x 572B's , directly grounded grids, metering via a W8JI article, swamped input, vacuum relay, HV supply with flux capacitor So, not normal in many respects. Hot spot on plate choke for 12m, arc welder. False dip on 20m nearly as you describe. The difference was as load was applied , arc welder. Did not like 3.92MHz. Would produce a very shallow dip and when load was applied, arc welder. As I attempted to do away with the goofy neutralization scheme Benton Harbor created because in part I directly grounded the grids and eliminated the HV choke input filter I found several more hot spots in the plate choke. In some cases one day it would be fine and others, arc welder. Then the "driver" , my TS-830S (or my backup Tempo 2020). It never happens on 160 (I just heat coax there at the moment) It never happens on 80, 40, or 20, 10m is a conundrum. So, with a tiny bit of neut/cap switched in I attempt to tune and load 28340. I have two results, both with near identical outputs. First result is RF on signal while watching the Ip drop out after a brief spike, poor tuning and single arc. Second is clean signal with consistent Ip and no arc but I barbque the 47ohm suppressors in the plate lead. Random result is near full output with only occasional arc welder without any warning. Now most here know I'm a self taught and mostly effective "hack" bordering on just enough engineering experience to be dangerous(lethal?) So with the help of the reticent reluctant reprobates here and elsewhere I found remedies. Swamping the input does not remove the need for neutralization. It just changes the value a bit and in some cases moves the capacitance to another place. Regardless of how evil the neutralization scheme is for the likes of the HA-10 or the 30L-1 it is still necessary but the tank tuning can not affect the neutralization capacitance.Refer to "arc welder" above. Both the Heath and Collins above are poor examples of how to accomplish neutralization, however within specific limitations they both work, marginally. Old components. First time I tried to load the HA-10 with 572B's installed, arc welder. Transformer, oil filled, choke, plate choke, door knob, and the 866's . I've learned a lot in 30 years My first failure with the TS830S was similar. I had a screen lay over and take out the screen and cathode resistors, half the HV, both tubes, plate choke, (rewound by hand) and doorknob/blocking cap. This led to knowing K4EAA (shameless plug) and K9TW (likewise). Although the TS830 is a robust and well thought out design it will not suffer fools gladly. A neutralized and robust tank circuit on any pair of 6146's will usually(usually) accept VSWR of up to 3:1 without hesitation on one condition, neither the plate or tune can be at maximum rotation in either direction, AND you must have a definite plate dip. If you can't find a dip or the plate and tune are maxed in either direction you have other issues and are about to arc weld. Again, I'm about a happy idiot when it comes to these things and I did state this was all anecdotal in nature and possibly more closely related to arc welding than Radio frequency engineering. My antique hardware talks to me and I listen. For now I get to talk more and weld less because I pay attention. Oddly what has saved me more times than not is a Drake MN2700B tuner. It absorbs most of my antenna and feedline mistakes and reliably provides a minimal VSWR to any radio or amplifier. Extra or not it's when I think I know it all that that I have problems. It's one of the reasons I hang around here.