I don’t have any special knowledge, so the following is speculation on my part; but it is how I understand events.
Bob Stuart’s technical papers and his Q&A on CA make it clear that MQA decoding could occur in software and even include the A/D D/A time correction, provided the A/D converter and D/A converter are known.
Initially MQA led Roon and Auralic to believe that they could implement software MQA decoding. They were advertised as MQA partners at CES 2016, but during the running of the show MQA and Auralic announced that Auralic would not be implementing firmware MQA decoding in the Aries, as “final rendering” or “the definitive version of MQA” required a hardware decoder in a certified DAC.
There was broad speculation (which I believe) that this change of position by MQA was commercial rather than technical in that royalties from dozens of hardware manufacturers were likely to be more lucrative than two software decoders.
Since then, Roon has been in negotiations/discussions with MQA as to software decoding in Roon. Roon is in a good position to do this, because (generally speaking) it knows what the DAC is. No announcement has been made by either Roon or MQA about the outcome of those discussions.
If the removal of Roon from the MQA site was recent then it may be an indication as to how those discussions are faring. If not, it could just have been the site catching up with the position post CES.
Given the time that has elapsed since the discussions started and the absence of any further information from either company I am not optimistic about the prospects for software decoding. If one took the (cynical) view that MQA was making decisions on a commercial basis (as they are fully entitled to do) then software implementation might occur only after the market for licences to hardware manufacturers was saturated.