I don't think the arguments for meta-tags being bad from Jeff's post apply to these tags.
The reason meta-tags are a problem is that they do not describe the content of the question.
reference-request does describe the content as well as soft-question. We haven't had any problem with them and there are quite useful. So I don't think there is any reason to remove them. I agree that there has been lots of bad meta-tags (e.g. subjective) but it doesn't mean that all meta-tags are bad.
How can you tell you’re using a meta-tag? It’s easier than you might think.
If the tag can’t work as the only tag on a question, it’s probably a meta-tag. Every tag you use should be able to work, more or less, as the only tag on a question. Meta-tags, like [beginner], [subjective], and [best-practices], are useless by themselves — they tell you nothing at all about the content of the question.
If the tag commonly means different things to different people, it’s probably a meta-tag.
We have had questions which were tagged just as soft-question. So I think what we call a meta-tag is different from what they call a meta-tag. We use meta-tags mainly for soft questions, i.e. question about TCS not in TCS. In comparison, non-meta-tags on cstheory are used in a similar way to subject classification and keywords for papers.