Since you seem interested in both user experience and measurements, I would highlight the following:
1) The CX-A5200, at the same output level (2.7 V) measured much worse if XLR outs were used, -72 dB THD+N vs RCA's, at -88 dB, that's 0.0251 vs 0.004%, a huge difference for people like Movie2099.
2) At reduce output level (2.4 V), the AV7705 measured -92.1 dB THD+N, that's more than 20 dB lower than the CX-A5200. We don't know how the AV7706 would measured yet but it is a reasonable assumption that it would be worse than the AV7705, though you never know.
So based on measurements I would conclude:
A1) CX-A5200, using XLR out will most likely be much worse than B1, a little better but still worse than B1 if you use RCAs.
A2) It doesn't matter because the 5200 would be the bottleneck.
B1) It would likely be the best way, again there is no advantage of XLR out except when the interconnects are long, and/or you are matching it with certain power amps in terms of gain matching, and in gain difference/or same between using XLR and RCA). In this case, the M33's gain is >29 dB so even if the gain using XLR or RCA are the same, you will not likely benefit from the higher output voltage. For many power amps such as Marantz and Yamaha's the gain for XLR input is 6 dB lower when using RCAs so the 2X output voltage would make no difference at all. Note: ASR seemed to suggest (based on measurements iirc..) the M33 would digitize the analog input signal. The Marantz would of course do that too unless you use direct mode.
So if you go with measurements, take the AV7706 if you must have an AVP/C, otherwise fyi in case you don't know yet, of all the D+M AVP/AVRs measured by ASR, any of the Denon AVRs measured measured much better than the AV7705, the only Marantz model that came close was the SR8015.
I am trying to address you question with my EE hat on. As to user experience, my AV8801 seemed to sound exactly the same as my AVR and my separates so I am not qualified to comment on the subjective side of things.
By the way, it is also a fact (based on many studies) that some people do prefer higher distortions so its is entirely possible that 1% distortion could sound better than 0.00001% depending on the distortion profile.