Wow! Bad caps for sure. Need better ones! The resistance of the cap would allow DC to pass, tending to bias one ring on relative to the 38 kHz signal, and the other off relative to that signal, which would explain the different 38 kHz residual readings at the grids of V102. Usually, excessive 38 kHz in the audio output points to the trap and de-emphasis networks. Have you changed any of the values in those networks? Checked the cap values? One sure fire way to test those networks is to swap the ring outputs at the grids of V102. It will switch the channels, but not the components. If the problem remains at the same output point, then de-emphasis is the problem. If the problem stays with the channel signal, then re-examine what the residual is at the two grids of V102.
Dave