Or that they have some other reason to have not learned as much (say, worse primary schools? differing cultural values?).
It is very well known that affirmative action tends to take the form of lowering standards for the groups being favored. If this is in fact true, running into a non-biased filter would tend to undo some of its effects. As noted towards the end of the post you're calling offensive (We also need to remember that many universities have quotas for minorities and give extra points on the SAT for just checking a box.). If this is not true... a weed-out class would not be observed to change the class demographics.
For instance, students from wealthier backgrounds are more likely to have had serious prior exposure to computer science before coming to Stanford, such as through high school computer science classes with well equipped computer labs. Students from much poorer backgrounds might not have had these pre-college opportunities.
This can make their first CS class harder for them--especially when it is a weed out class at a top CS school.
So you might have four kinds of students.
1. Those who had significant CS experience before Stanford, and came to Stanford with the intent of majoring in CS.
2. Those who had significant CS experience before Stanford, but came intending to major in something else, but are taking the intro CS course to see what CS is like.
3. Those without a significant pre-Stanford CS background, but came intending to major in CS.
4. Those without a significant pre-Stasnford CS background, and came intending to major in something else, but are taking the intro CS course to see what it is like.
Those in group #1 will have the best chance of not getting weeded out.
Those in #2 will also have a good chance, because of their background.
#3 is more iffy. Their school background lacks CS, but they intended to be CS majors, so many of them will have self-studied CS to make up for their poor formal education.
#4...these people are going to have a seriously hard time. They are much less likely than the #3 people to have done anything to make up for their high school's inadequacies in CS.
I'd expect the distribution of minorities among these four groups to skew towards #3 and #4, and so I'd expect them to get hit more by the weed out class. This is not because they are not as smart--they are just not as well prepared coming in due to inequalities in pre-college education.
TL;DR: the pipeline problem doesn't start at college.
Say the course is very hard and it cuts 2/3rds of the total students. If the cuts were proportional by race, we now have 6 white, 4 asian, and 0 blacks.
There was no racism involved in the class, the fact that no blacks remain is a result of the class brutally weeding out 2/3rds of the students and that the initial number of blacks was very low.
You can quibble over numbers, but the basic idea remains that large percentage cuts in an unbalanced population can make it tend more toward homogeneity.