Let $L$ be a first order language. Let $T$ be a set of sentences in $L$ and $S$ a sentence in $L$. Let's define a meta-proof to be a proof that there exists a formal proof of $S$ from $T$.
Question: Are there sentences in the "known" mathematics whose meta-proof is known but whose formal proof is not known.
Remark: when we add constants/defintions to our language and prove the sentence in the new language, we know it works because there is a theorem that says that under suitable conditions a formal proof in the new language implies the existence of a formal proof in our original language. Therefore such a situation counts as an instance of a meta-proof without knowing the formal proof, but I would consider this trivial.
Thank you
Parikh sentences have a short proof that a proof exists but the actual proof can be very long, for example the number of symbols could be greater than the number of atoms in the universe. For a discussion see page 17 of A Universal Approach To Self-Referenctial Paradoxes, Incompleteness And Fixed Points by Noson Yanofsky. This paper is quite readable.
Paraphrasing this paper uses a diagonal argument to construct a sentence ${\mathcal{C}}_{n}$ which says "I do not have a proof of myself that is shorter than $n$."
There follows a short proof that there exists a proof of this sentence.