Let's say a new ASIC comes out that mines twice as efficiently, i.e. 2x less energy per calculation. Because miners can now afford to run twice the amount of ASICs on the same energy budget, this means that eventually miners are mining twice as fast.

The network however needs to keep the block rate constant (6 blocks per hour in Bitcoin), so it counteracts by adjusting the difficulty. In this case, mining a block becomes twice as difficult. In summary, everyone has now twice the amount of ASICs consuming the same amount of energy, still mining at 6 blocks per hour.

The misconception is that the "work" in "proof-of-work" is meaningful by itself. It is actually not, it is a lottery whose probability to win is adjusted exactly so that statistically, all miners in the world together will mine 6 blocks per hour on average.

As for "mining at a loss": I am not convinced there is a general principle that only allows for absolutely clean energy to make mining profitable, across the whole world and in any situation. We have already seen a coal plant being ramped up for the purpose of mining in New York, and hash rate going down when coal plants in China got flooded.