There's this principle, which is good to apply when you can; and then there's the principle of choosing your audience. If you explain a fact in plain language in one sentence, you will miss some percentage of skimmers. If you bring it up four times, you will catch more skimmers and lose anyone who wants a faster pace and less repetition. Similar balances hold for what things do and do not cost suspension of disbelief. If you obey the reader who finds Randi to be a challenge to that suspension, then you weaken your hold on the reader who thought the original version of the tidbit was charming, and has never heard of this replacement fellow. And the reader who disapproves of excessive editing after the fact.

I shouldn't need to explain this to you. You have authored essays on the concept of 'subjectively objective' and my statement was quite clear and even noted that it was purely a technical tangent. In fact, the upvote on its parent was mine.

"What the reader says" is not the same thing as "what happened to the reader". 'What happened to the reader' can be fully determined by the timeless state of reader themselves but is not necessarily the same thing as what the reader says. Just as someone who says "my prior for A is 0.34" when their prior is actually "0.87" is wrong, despite the fact that priors are subjective. Subjective does not mean what people say about themselves must be true.

but if the reader says you're messing with their suspension of disbelief, the reader is always right.

Still false.

If that's what happened to the reader and broke their suspension of disbelief, that's what happened. It doesn't matter if the reader made a mistake. Your text caused that mistake.

True.