Submitted by ulqiorra, also found by joestakey, 0Kage, and 0xMirce
https://github.com/code-423n4/2023-01-popcorn/blob/d95fc31449c260901811196d617366d6352258cd/src/utils/MultiRewardStaking.sol#L406 
https://github.com/code-423n4/2023-01-popcorn/blob/d95fc31449c260901811196d617366d6352258cd/src/utils/MultiRewardStaking.sol#L427 
https://github.com/code-423n4/2023-01-popcorn/blob/d95fc31449c260901811196d617366d6352258cd/src/utils/MultiRewardStaking.sol#L274
Reward deltaIndex in _accrueRewards() is multiplied by 10**decimals() but eventually divided by rewards.ONE (which is equal to 10**IERC20Metadata(address(rewardToken)).decimals()) in _accrueUser().
If the number of decimals in MultiRewardEscrow share token differs from the number of decimals in the reward token, then all rewards are multipled by 10 ** (decimals() - rewardToken.decimals()).
Therefore, for example, if an admin adds USDT as the reward token with decimals=6, it will result in the reward for any user to be multiplied by 10**(18-6) = 1000000000000 on the next block. This will at best lead to a DOS where no one will be able to withdraw funds. But at worst, users will drain the entire reward fund due to inflated calculations in the next block.
Put the following test in ./test/ folder and run with forge test --mc DecimalMismatchTest. The test fails because of incorrect supplierDelta calculations:
Use the same number of decimals when calculating deltaIndex and supplierDelta.
RedVeil (Popcorn) confirmed 
