Submitted by WatchPug, also found by Czar102
HolyPaladinToken.sol#L715-L743
When current time is lastDropUpdate + (2*MONTH-1):
nbMonthEllapsed will be round down to 1, while its actually 1.99 months passed, but because of precision loss, the smart contract will believe its only 1 month elapsed, as a result, DropPerSecond will only decrease by 1 * dropDecreasePerMonth.
In another word, due to the precision loss in calculating the number of months elapsed, for each _updateDropPerSecond() there can be a short of up to 1 * dropDecreasePerMonth for the decrease of emission rate.
At the very edge case, if all the updates happened just like the scenario above. by the end of the dropDecreaseDuration, it will drop only 12 * dropDecreasePerMonth in total, while its expected to be 24 * dropDecreasePerMonth.
So only half of (startDropPerSecond - endDropPerSecond) is actually decreased. And the last time updateDropPerSecond is called, DropPerSecond will suddenly drop to endDropPerSecond.
As the DropPerSecond is not updated correctly, in most of the dropDecreaseDuration, the actual rewards emission rate is much higher than expected. As a result, the total rewards emission can be much higher than expected.
Change to:
Kogaroshi (Paladin) confirmed and commented:
