There's plenty of double blind ABX test results at https://hydrogenaud.io that contradict this, and with some material at even higher bitrates. Mostly it's experienced listeners making useful contributions to codec developers.

The worst thing we ever did for audio quality was focus on short A/B tests. It seems like a decent idea, and works for big differences, but ears are not great at it.

Much better results would be had with 1 month of A, then switching to B.

Any sources?

I can’t find the references I got that from (it was decently respected people in the audiophile world talking about their experience with AB tests over a decade or two), but it’s also true in my own experience.

As well, I feel like the same concept is being used when people listen to the same recording of the same track every time they are trying out new headphones or speakers.