https://github.com/code-423n4/2022-12-forgeries/blob/main/src/interfaces/IVRFNFTRandomDraw.sol#L71-L90
https://github.com/code-423n4/2022-12-forgeries/blob/main/src/VRFNFTRandomDraw.sol#L249-L256
https://github.com/code-423n4/2022-12-forgeries/blob/main/src/VRFNFTRandomDraw.sol#L271
Accordingly to EIP-721:
This project, aims to create a raffle specifying that the potential winner will be between a range, where the lower limit is set by the candidate with the lowest TokenId and the candidate with the highest TokenId (+1 to be included in the draw) sets the upper limit. As stated in the previous quote, this could generate gigantic ranges with numerous empty tokens given how it is calculated (see ENS as an example of empty slots and how the ids are generated).
After generating the random number via VRF, the winner is selected by moduluing by the range plus the initial token id. The result is then used to determine the winner. 
Considering the costs to query VRF and the waiting time to claim the prize, this issue may turn the contract unusable. 
Note that any used method when casting the collection to IERC721EnumerableUpgradeable has the same effect as casting to IERC721, and giving how a raffle is setup, it seems the original intent was to use indexes instead of ids.
External requirements:
Consider a scenario where:
Setting a raffle to include each possible token results in [1..1001[ alternatives. Since there are only three possible winners, theres only 0.3% of a successful draw.
Depending on the direction the project takes:
