Well, one for video and one for audio....
So no, I think there is only one decoder (ie module which decodes the M2TS into separate PCM video and audio streams).
The 7405 SoC has serial inputs (PKTn, each comprising Data, Clock, Sync, Valid, Error) through which the demodulated DVB multiplex streams from the CXD2820R tuner demodulators must be connected. I suppose that 2 of the 6 available PKT inputs are used, 1 per demodulator. The demodulators in turn are managed by I²C, presumably using the SoC's BSC I²C test interface.
The streams are received, buffered, demuliplexed and (if necessary) decrypted by the SoC's Data Transport module according to the selected programs (PIDs), passed to the Video and Audio Decoder modules, and then to be displayed/recorded/RF modulated. The part of the installed memory that isn't visible to Linux is presumably being used to stage this process.
As to the fault in question, if the fault is affecting both tuner/demodulator paths, the SoC internals would seem to be fingered, with two or more separate bad ball connections to the chip as a lesser possibility. Or just possibly the PSU is unstable while warming up, sufficiently to cause the tuner/demodulator paths (eg) to fail without crashing the SoC; perhaps there's a separate rail for the tuners that's slow to come up?
Clearly scheduling two recordings to start simultaneously on different muxes will show if both tuner/demodulators are affected.