Submitted by simon135, also found by 0x4non, ForkEth, zaskoh, and ArmedGoose
https://github.com/rabbitholegg/quest-protocol/blob/068d628f019e9469aecbf676370075c1f6c980fd/contracts/Quest.sol#L113-L116
If the reward token is erc1155/erc777, an attacker can reenter and then buy/transfer another unclaimed token to the attacker address and then the var redeemTokens wont be equal to how many tokens were actually redeemed.
Example:
Reward token is an erc1155 that has  _afterTokenTransfer
Alice(attacker) has 2 receipt tokens, the first one is on a smart contract that will do the reentrancy, and the second one is on Alices address but is approved to transfer to the smart contract(the own that holds the first receipt)
Result:
The invariant that redeemedTokens = tokens that are redeemed is false because it doesnt account for the first token that we reentered.
The issue is worse  with erc777 tokens because of  the fact  that accounting will be in  the withdrawRemainingTokens function
after the reentrancy
ex: redeemedTokens=9 but should be 10
receiptRedeemers()=12
rewardAmountInWeiOrTokenId=1e5
unclaimedTokens=300000
assuming they are some tokens left
balance(address(this)=201000 and protocolFee=500
nonClaimableTokens=201000 - 500 - 300000 it would revert ( negative numbers  with uint) and   funds would be stuck in the contract forever
The real estimate for nonClaimableTokens=201000-500-200000=500 and the owner can get funds out
but 500 wei will be lost in the contract
and  it can get worse with large amounts of quests and the attacker reentering multiple times to cause a bigger gap between the real redeemedTokens
Add nonReentrancy modifier
kirk-baird (judge) decreased severity to Medium
waynehoover (RabbitHole) confirmed
