[Amps] Really silly question

[Peter Chadwick]

So I've had this problem appear on the amplifier......
Bear in mind I've had chance to see it, but not chance to dig in to find out what's happening, and it's a fault that has appeared after some 19 years, so is arguably not a design problem.
Amp is a pair of passive grid, parallel 4-250A, 300 volts regulated on the screen, shunt regulated bias, 3700volts on the plate,  with an interlock that kills the power when the bias dies.
The input is a 4:1 impedance transmission line xfmr wound on a 2 inch diameter iron dust core, with a shunt load consisting of 4 off 50 ohm, non inductive wire wound 50 watt metal clad resistors seriesed up to give 200 ohms. I can't remember the transformer inductance.
Running into a good 50 ohm load, all is well at 3.5MHz. At 3.75 up to past 4 MHz, tuning the PA to resonance leads to a peak in PA current, and the bias apparently collapsing, leading to the amp tripping off - just as if it was oscillating.
As I say, I haven't had chance to dig into it. First thoughts are that maybe the the resistors (or one of them, anyway) have gone open, and the input transformer is resonating around 3.8MHz with the input capacity of the tubes. Interestingly, on 1.8, 7, 10, 14, 18, 21 and 24 MHz, the amp works OK. I haven't checked with any detail the gain, which if the resistors had gone open, would be expected to go up. Output tuning is big ceramic caps for tuning and loading, a variable inductor and a 500pF variable on the output in shunt with a set of switched capacitors. The same tuning cap is used across the whole 80m band.
And as I'm travelling this week (again!), I can't readily go and look.
Any ideas? And I don't readily accept VHF parasitics........the shunt resistors are 25 watt dissipation Carborundum types, solid carbon rods, and as far as I know, still at the original value.
73
Peter I3/G3RZP