[[selaphobia]]
Therein potentially lies a problem: these words go together as a semantic block. The condition is selaphobia, the description is selaphobic, the person suffering is a selaphobe. This is all perfectly regular, to the extent that if we can attest the concept, we may as well use that attestation to work for all (trivially predictable) forms. It's not exactly the same situation of using the genitive or plural as attestation of the lemma, but it is, I submit, very close. (We might even extend the idea to a template of related terms, such as "autism:autistic:autist", "schizophrenia:schizophrenic:schizophrenic", &c.)
Otherwise we've got the situation that we might not be able to find three attestations of "selaphobia", but can find only one each of "selaphobia", "selaphobe" and "selaphobic", and therefore the concept itself must be deleted in its entirely, being unattestable in any specific form.
As another question, all these -phobia words are largely from word lists... but they are in a lot of those lists, including fairly official-looking ones. (Medical dictionaries and the like.) Is it worth including these terms anyway, with a label or etymology note that it's (almost) exclusively found in such a list? Because I can see someone using such a word (because it's commonly packaged up in lists for that very purpose), someone else wanting to know what it means, and when they come here they can't find out because we've decided that it doesn't really exist. When we could have the far more useful "this word means "blah" in theory, but doesn't seem to be used in practice. It was probably invented for lexical completeness' sake."
We have such a list: Appendix:Invented phobias. I've mentioned it to the creator of these recent unattestables.
If you have to know exactly where to look to find a term, because it's not otherwise findable, maybe it's not so useful to hide things away there?