We are again stuck at point one. MQA is not LOSSY in any way you are using.I mean, this is the same discussion that happens with any lossy compression: How much (and what kind) of the relevant information can be retained while compressing the filesize?
You may argue that it's "perceptually indistinguishable from lossless" in some contexts, which is a separate argument - but either way, if the compression can't be reversed to the original, it isn't lossless.
CD release from a modern DXD master is also LOSSY by the same definition - you can't get back to the master.
That's why John Atkinson tried to reframe the question of MQA lossyness by asking "compared to what?"