> Not eating meat however is a strong signal though

I also personally know several counter-examples to this rule.

If meat/no-meat is the only signal people are using, there must be a lot of false positives and negatives. Or my sample of the population is somehow atypical.

Yes it is not, plenty of soft cues are signals - today's generation isn't good at identifying them ( a good thing!).

Meat/no-meat is the most common signal, certainly not the only one, surname would be another, language (ethnolect or dialect) is sometimes another.

There are no perfect methods short of asking, some older people and more overtly racists (even they aren't aware) straight up do ask.

Also most racists only really care whether you are their caste or not, it doesn't matter what you actually are, so classification into their group or not is easier than do it precisely.

There are a lot of false positives and negatives, which is why this is a broken system in cities and places where you couldn't possibly know their caste from their diet or their surname. Nothing short of explicitly asking for their caste would let you know 100%, people are just really good at generalizing.