Hi!
I have had many problems with germanium-transistor O/P stages over the years, both with getting the bias right and preventing thermal runaway!
What you can try is the following:-
1) Obtain a pair of BC327 medium-power output transistors from fleabay, etc (Hong-Kong SS8550 etc., will also do!), and twist the base/emitter leads together, connect a thin length of fine insulated connecting wire a few inches long to the paired leads, solder and sleeve the joint with thin heatshrink sleeving;
2) Solder a second length of similar wire to the first to the collector lead of the transistor, solder and sleeve the joint as before in step 1) above;
3) Obtain some Thermal Bonding Adhesive, apply a SMALL dot of adhesive to the top of the O/P transistor's case Q302, secure the Silicon Bias Transistor prepared in step 1 and step 2 above and leave to set;
4) Repeat Steps 1, 2 and 3 above with the second O/P transistor Q303,
5) Obtain a couple of 1.0 R 2W resistors, (these will take the place of R307 & R309) soldering ONE END only to the circuit-pad connecting to the junction of the driver transformer secondary winding for each half and the 1k 2W bias resistors R308 & R310 - see attached circuit-extract;
6) Connect the FREE end of each 1.0R 2W resistor R307 and R309 to the lead coming from the SINGLE COLLECTOR LEAD of the Bias Devices fitted in Steps 1 to 4 above to each O/P transistor case;
7) Connect the lead coming from the PAIRED EMITTER/BASE LEADS of each Bias Device fitted in Steps 1 to 4 to the original connection pads of R307 & R309 that connect to the +31V power supply and the LS mid-point of the O/P stage respectively;
DOUBLE-CHECK YOUR WORK!
9) Connect a DVM, set to a mV range, across R313 and switch on. A reading of no more than about 6-8mV should be obtained - if the current starts to rise rapidly, switch off at once and double check your work - you may need to fit a lower value in place of the 1.0R resistors fitted in step 6 above;
10) What you are aiming for is to get a current across EACH emitter resistor as close to the maker's value as possible - if Vox didn't give a figure in their Documentation, try for about 4-6mV across each emitter-resistor to begin with!
11) The Silicon Bias devices will have a negative Vbe/Vcb temp. coefficent of voltage approximately the same as the replacement O/P Devices you fitted, so some degree of temperature compensation should be obtained.
The above took me far longer to write and explain it than it does to actually carry it out, believe me!
The circuit-extract should make it clear, I hope!
Chris Williams