The lead dress is the key, and
IME it will be the dress around the phase inverter and
output tubes that will make or brake you. Be carefull
testing while the amp is oscillating - you can overheat
the power tubes without hearing much at all (they can
be going ballistic at a very HF). One thing to be carefull
of are the primary leads of the OT - keep them tight
together and away from the board (tight down against
the chassis). The secondary leads should be as short as
practical and away from the pre-amp leads as much as
possible. 
Some little things that *might* help:
 
1. if you don't already have 5.6k grid stoppers, now's a
good time. 
 
2. try a bigger cap across the PI plates (stock is 47pF,
try 100pF or 250pF) Maybe not something you want to
leave in there, but it can help find the problem in the
meantime. The longtail PI is prone to a self osc around
47kHz if I recall correctly.
 
3. Try putting a 100-ohm resistor on each power tube
socket on pin 3. Attach the plate lead to the resistor.
This can help dampen an osc in the plate circuit (by
changing the Q). It won't affect the tone.. no worries. 
 
4. Shielding the input jacks sometimes does the trick.
(you have to fabricate a metal shield to go around the
jacks.. a good bit of work.)

You might VERY CAREFULLY
poke around the wires leading to the preamp tube
sockets, especially the (generally) purple wire that
carries the negative feedback signal from the impedance
selector to the circuit board. I like to use an insulated
pair of small neddlenose pliers for this. 
 
I would set the amp up so it is oscillating and then start
CAREFULLY (did I stress that enough?) moving wires
around. If the oscillation changes, you're probably
pretty close. I would then try to find a wire position
where the amp doesn't oscillate. 
 
In general, I usually put the filament and grid wires as
far apart as possible and I do this by pushing the
filament wires against the chassis and leaving the grid
wires up in the air. You have to be careful about where
they go though.