I am making a protocol that has one missing piece. It involves 2 non-trusted party $A$ and $B$ and a semihonest third party $C$. This protocol has a contingency step where if $A$ leaves prematurely, $B$ is supposed receive time locked puzzle. It is important that $B$ cannot get the puzzle early. The puzzle ensures that after a period of time, $B$ will get a value he is supposed to get, but not immediately. Neither $A$ nor $B$ want $C$ to be able to learn the value hidden in the puzzle (therefore $C$ can't have the puzzle), and $B$ shouldn't be able to start working on the puzzle until $A$ leaves and $C$ reveals the puzzle. $A$ also needs to prove to $B$ that the value hidden in the puzzle matches a commitment (preferably Pedersen) at the beginning of the protocol. Note that once $A$ leaves, $C$ wishes to leave almost immediately (for important structural reasons), so it can't just be an encryption to which $C$ holds the key and $C$ gives it to $B$.
My question is does such a system currently exist? This system being as follows: $A$ creates a time locked puzzle that cannot be worked on by $B$ until a value is released while still proving that the puzzle hides the same value as a commitment.
I know that such a scheme can be created by going down to the bit-level, but I would strongly prefer a more efficient solution.
I am aiming for a sequential puzzle that takes around 2 hours for a powerful computer to solve, akin to this paper.
TLDR: Is there a way to encrypt a puzzle such that the puzzle cannot be worked on until it is decrypted while still proving that the solution to the puzzle matches a commitment?