The script allows the remote node to claim funds with the payment pre-image and the local node after a HTLC timeout. The author points out that the output that sends the funds to the local node is not timelocked in the script, but it is delayed in the second stage using to_self_delay. However, the author believes nothing stops them from offering an HTLC to their peer node, signing the updated commitment transaction, and broadcasting the commitment transaction on the mainchain to spend the HTLC to an address locked with CSV. The remote node would still forward the HTLC because it is within the bounds of the cltv_expiry specified by BOLTs. The author questions where the shortcoming is in their thought process and asks if timelock expiry should be included in the script itself. The author has already asked this question on Stackexchange with no response yet.