In college, one of my friends took a philosophy class where the professor didn’t adjust the grades based on which of the two TAs you had (they were responsible for grading the papers). One of the TAs was much tougher than the other one, but the professor’s reasoning was that people didn’t know this going into the class, so they had as much of a (random) chance to get the hard TA as the easy one. Because it wasn’t ex-ante unfair, he saw no need to adjust based on ex-post TA harshness.
At first, I thought the reasoning was kind of ridiculous. If you see that the highest grade one TA gave was a B and the other one gave out A’s to 60% of the class, do you really want to argue that nothing should be done about this outcome because of ex-ante equal probabilities of getting the harsh TA? Sure, you’ll be "redistributing” some of the good grades away from the people with the easy TA, but wouldn’t they have wanted that if they didn’t know which one they would get? By the way, this also happens with professors teaching the same class at very different difficulty levels and not adjusting for that when they give out grades.
Of course, the story can be a little more complicated. What if the "easy” TA’s session was at 8am and the "hard” TA’s session was at 2pm? Could it be that the TAs are the same, but the people signing up for the 8am session were smarter than the people who wanted to sleep in? And what if people knew about the TAs’ reputations going into the class?
In this case, of course, you can solve the problem somewhat by having the TAs grade random papers, not just their own students’ and then de-meaning the grade by each TA’s average. Or you can have them both grade each paper (not like they have anything else to do) and take the average grade. Or you can find out which one the easy one is and sign up for her section.