I was not surprised by this reaction. Alkyl migration is noted in Williamson,
Macroscale and Microscale Organic Experiments.
I am not going to be a sympathizer on the mechanism issue. Although I too favor teaching reaction mechanisms, I am reluctant to place too much truth to what may be written. I expect that between initially learning how electrons move to form a product, the description(s) are not all inclusive. That is the case for this reaction. The ortho/para products do form initially, but rearrange to the meta-isomer. If there was a failure, it would have been a failure to rationalize this reaction. It is useful to attempt to rationalize formation of unexpected products, not necessarily to learn an unusual path, but rather to recognize competitive elements present in reactions. A simple example is a Claisen condensation with an alkoxide. Many students forget that extensive attack is also occurring upon the carbonyl group because it is a non-productive reaction. If a different alcohol were used, ester exchange would be much faster than condensation.
I too was schooled in a rearrangement reaction related to a
pinacol rearrangement. I had not expected any rearrangement to occur as I was unaware of by-products in carbonyl group protection reactions. However, it seems that rearrangements can occur for some compounds when an alcohol is not present to trap the intermediate protonated carbonyl group. It isn't necessarily a failure of the mechanism, but a failure to recognize how the intermediates might react in the presence of other potential reacting groups. Did you know benzene-d6 is made with benzene and D2SO4 (I think that is how it is done)?