The pre-strike current flow certainly was stopped by the grounded grid. Good. Upon striking, my instruments showed 4 mA through the tube and somewhere under 2kV dropping across it. That was just as it should be. There was no flicker on and off as in the initial experiment. The main problem, however, came again from a high-pitched noise accompanied by the cooling fans' stoppage. The thought of the anode seal overheating and the oscillation destroying my FETs led me to shut down the experiment after 10 seconds or so.
So, from whence comes this oscillation? If I had an oscilloscope (seems obvious, right?), I think I would probe (1) the op-amp output that is driving the MOSFET or (2) the FET drain. The suspected candidates include:
(1) An unstoppered triode anode, grid, or cathode could cause the oscillation.
(2) An academic paper I came across suggested that current regulator feedback chains should have a limited response to around a couple kilohertz in order to avoid "severe oscillations". I tried an integrator on the gate drive op-amp, but the same behavior was observed.
The oscillation's frequency might interfere with the +24V PWM feedback loop or the DC fan controller; however, this behavior has been observed when the +24V is supplied by batteries. So the former seems unlikely.
Could it be that in input voltage presents its own spikes and noise that get multiplied in the PSU? I'll try to cause this behavior again, then work on muting the power supply spikes, if found.