I guess it's a typical big company nonsense. Someone probably had an idea that daily cleaning is too time-consuming and expensive, and the task to solve that problem was introduced. Some lab made tests, and defined The Only True Way of keeping the product edible. (Side note: re-boiling a mix of fats and sugars each day for two weeks? Who would eat the result of that? Who thought it was a reliable idea?) The cleaning period could only be stretched to the maximum, for obvious business reasons. That resulted in tiny error margins (which maker can't voluntary change without spending money on new research). So presentations on clear financial gains were made, and the new option was proudly implemented. Some employee has probably been keeping eye on changes in cream composition, regulations, and related research, and issuing new checks for the machines. However, in reality this doesn't work that well (even indifferent part-time workers who don't give a damn after a shift are part of reality you can't simply rule out). But the goal of reverting or altering the process is not really important to McDonalds, as making it work is now pain in the store owner's ass.

You have probably noticed similar cases in computing, when some high-tech solution are introduced, and their gains are shared with lots of pathos, but the resulting system looks way too complex and unstable even from a distance.