I couldn't find the stage designation in the article, so I assume that Stage 1 = No Decode; Stage 2 = Decode to MQA Core at 88.2/96kHz; Stage 3 = Full Decode.
Actually the most important difference between MQA Core signal at 88.2/96kHz and the full decoding (includes rendering) is this: the DAC-specific optimizations and whatever filters MQA apply in the rendering. I believe the final sample rate does not matter as much since 3rd party tests show that there's not much musical information above 60kHz.
By the way, people always ask for a software solution to do the full decoding without buying new hardware, that's just not possible because the rendering is tuned for individual DAC hardware. If a DAC manufacturer does not ship a DAC to MQA Ltd. for certification, there's no tuning data for the software to exist. If a DAC manufacturer ships a DAC for certification, then it gains the MQA decoding or rendering functionality, so there is no longer a need for the software to do full decoding.