Word. The 1990's CS50/CS51 course enrolled both students who'd gotten a "5" in the CS AP and students who didn't know what a for loop was. It felt like learning to swim while simultaneously drowning.
The even sillier split-audience course in the mid 2000s was CS121, the intro CS theory course.
More or less a math course, but required for CS students, the audience included a mix of advanced math students (including e.g. some IMO winners) and programmers without any math background.
The result was that half the class felt it was incredibly easy and slow-paced (at least for the first month; later the problems got tedious and fiddly for everyone), and the other half was completely overwhelmed.
As referenced in sibling comment, I took 121 a bit earlier than that, with Prof. Lewis. Of course he taught from his "Turing's Face" textbook, which is widely touted as accessible to students with high school math. By the time I took 121 I had quite a bit more math than that so I can't recall whether that is true. I agree with your "tedious and fiddly" assessment, but I don't see any way around it. CS is a tedious and fiddly subject anyway, but the fiddly tedium in this case is related to foundational truths about computation rather than trivial details of particular algorithms (...or, at less ambitious schools, APIs). Frankly, I hope it's never the case that a student could graduate Harvard with a degree in CS (or applied math) without mastering the material that Prof. Lewis taught in CS-121.
CS51 in 1997 (to specify the version to which I was exposed) was a great deal of work, but it wasn't beyond the capability of the average Harvard student. After all, as mentioned ITT, the Great Pointer Winnowing had taken place in CS50. Maybe I'm speaking from privilege, since I had coded some BASIC in 8th grade...
