Submitted by t0x1c, also found by Banditx0x and 0xAsen
RewardsEmitter::performUpkeep() distributes the added rewards to the eligible staking pools at the rate of X% per day (default value set to 1% by the protocol). To ensure that once disbursed, it is not sent again, the code reduces the pendingRewards[poolID] variable on L126:
The impact of this is: higher the number of times performUpkeep() is called, lesser the rewards per day is distributed to the pools. That is, calling it 100 times a day is worse than calling it once at the end of the day. This is at odds with how the protocol wants to achieve a higher frequency of upkeep-ing. 
Reasoning:
This above code logic is incorrect maths as doing this will result in the following scenario:
So on and so forth for each new day. The actual formula being followed is totalRewardsStillRemainingAfterXdays = 10 ether * (1 - 2.5%)**X which would be 3632324398878806621 or 3.63232 ether after 40 days. In fact, even after another 40 days, its still not all paid out. Please refer the Proof of Concept - 1 provided below. In fact, since this is a classic negative compounding formula ( $Amount = Principle \* (1 - rate)^{time}$ ), no matter how many days pass, all the rewards would never be distributed and dust would always remain in the contract.
To remove any doubts regarding the interpretation of the 1% per day rate and if indeed the current implementation is as intended or not, one can look at the following calculation provided in the comments:
StakingRewards.sol#L178-L181
Lets crunch the above numbers:
This is certainly not the case currently as can be seen in the Proof of Concept - 2. 
Note that this means that per day much lesser rewards than 1% are distributed if performUpkeep() is called every 15 minutes instead of once after 24 hours. A malicious actor can use this to grief the protocol and delay/diminish emissions. The power of negative compunding causes this and can be seen in the second PoC.
Add the following test inside src/rewards/tests/RewardsEmitter.t.sol and run via COVERAGE="yes" NETWORK="sep" forge test -vv --rpc-url https://rpc.ankr.com/eth_sepolia --mt test_allPendingRewardsNeverPaidOut to see the two assertions fail. The value of rewardsEmitterDailyPercentTimes1000 in the test is 2.5% per day.
Add the following test inside src/rewards/tests/RewardsEmitter.t.sol and run via COVERAGE="yes" NETWORK="sep" forge test -vv --rpc-url https://rpc.ankr.com/eth_sepolia --mt test_perDayDefaultRateNotAdheredTo to see last assertion fail. The value of rewardsEmitterDailyPercentTimes1000 in the test is 1% per day and upkeep() is called every 15 minutes.
We observe that by the end of the day, 4931 ethers less is disbursed than expected.
Foundry
Store the eligible pool reward & the already paid out reward in separate variables and compare them to make sure extra rewards are not being paid out to a pool.
Irrespective of the performUpkeep() calling frequency, at the end of the day, 1% should be disbursed.
Picodes (Judge) decreased severity to Medium and commented:
othernet-global (Salty.IO) acknowledged
