This is classic collector emitter leakage resistance across a cascode stage before complementary emitter follower stages. The result is insufficient DC bias for just one polarity .
It would be on the high side PNP transistor that leaks starving the low side of DC bias so they do not overlap at 0V. Diode tests won't work.
Unfortunately without a schematic , I can't give you the Ref Des.
but if you put your finger across the collector emitter of the "right" part with a light touch (100k) you can make the signal worse and thus replace it. There should be a bias pot to null crossover distortion, but that may not work.