R26 and R27 values sound plausible if you interpret as 2.53K and 0.08K ie 2530 ohms and 80 ohms. That would give around 0.275V base bias for the output transistors at zero base current. I would expect each OC81 to need about 200uA for (say) 10mA collector current which would drop the bias voltage by about another volt which would reduce it to <0. Which couldn't happen, so what it means is the standing current in the output devices should be somewhat less than 10mA which would be fine.
The description of the capacitor test on the DVM sounds fine- at the low voltage from the meter it charges up and the resistance appears to get higher and higher; when you reverse the meter it reads the charged up capacitor voltage as negative resistance until it has recharged it positively again. The capacitor is not leaking when charged backwards since the applied voltage is low. Someone suggested testing the capacitor by hooking it across the battery in series with the DVM on current which might expose some weakness that the simple resistance check doesn't, I think that would be useful to try.
Have you tried disconnecting the output transformer from everything else and looking for any leakage resistance between its separate windings- that could be a path for excessive current though it wouldn't make the output devices overheat.
A similar problem in the driver transformer could cook the output devices, however.
Keep trying, there's only a dozen components and some of them are already known to be OK. The answer will eventually seem obvious, but only after the event
Chris