Hi-Fi (let alone High-End) is a though topic, there is no ultimate solution, but opinions and preferences...
Relying only on THD measures may be misleading, as these parameters are born with some sort of static tests. To my understanding in the 70’s-80’s enthusiast engineers found out what TIM (Transient InterModulation) is. Until that amplifier designers focused on high open loop gain, and relied on the global feedback loop to decrease THD. Indeed this way it is possible to get very good THD measures for the borchure, but the trade off here is that the open loop bandwidth is low, which implies that the feedback loop „speed” is also bad. High TIM is the result of this suffering, and is often even noticeable when listening to music on such amplifiers.
Making a Low-TIM amplifier is a design concept, which relies highly on symmetrical circuit layout, and on more local feeedback loops rather than big global feedback.
There is plenty of information on the net over this topic, IMHO, this is the real game changer in the amplifier construction.
But this is only one aspect of the things, because a lot depends on actually what music do you listen usualy. Electrical music (Yello, J.M. Jarre, etc.) may sound very impressive on virtually any sound system, which may not be the case with a brass big band style music...
I enclose a mid 80’s amplifier schematic, which uses classic semiconductors but feature the „novel” circuit arrangement already, this amplifier was a part of Hungarian made active studio monitors, each amplifier drove a single speaker after active crossover network. It is a rather complicated circuit but if one takes the time to analyze can see each component have a good reason to be there...