[Argus25]

One good trick to avoid loading the oscillator under test is to clip the scope probe onto the outside of the insulation of a piece of wire in the circuit. It makes a fraction of a pF gimmick capacitor and if you wind up the scope gain you will still be able to see the oscillation (but the amplitude data will be meaningless) but at least it won't alter the circuit you are testing to any significant degree.

The fact it is bombing out at the HF end, means likely there are frequency dependent losses somewhere in the feedback circuit/pathway, assuming the main resonant circuit has not lost Q. Possibly a feedback pathway in the socket from plate to grid & phase cancelling the + feedback sustaining oscillations.

The main resonant circuit is easily checked at all frequencies if you break the feedback to stop the oscillator and feed a generator signal in instead.

Possibly mouse urine has contaminated the ceramic valve socket, that ionic compound won't clean off well with contact cleaner either, needs deionised H2O, soak for some hours, or replace socket.