Why are they redoing the rest of the course then? If the only change is the date then things are fine. However research does move on, if they need to redo the course materials for any reason that needs to include the lectures.
It still takes time and money to record videos or collect assignments and ensure legal compliance, privacy, accessibility, etc. which often utilizes student labor, where there's an opportunity cost. That is, students, professors, and everyone in-between could be frying bigger fish.
If a course is teaching computer architecture and switches from a hypothetical instruction set to RISC-V, the core of the instruction hasn't changed---memory layout, instruction ordering, digital system concepts.
(The nice thing, Silvina Hanono Wachman uploads a privately recorded version of each semester of the computer architecture class, so we get the classic Chris Terman instruction as well as the modern variant.)
With that in mind, some classes will have a higher relevance churn rate than others and should be updated more often. Other classes might have historically had an amazing instructor and no major development and thus need a really compelling reason to be replaced. (MIT's Introductory Physics comes to mind.)
> Other classes might have historically had an amazing instructor and no major development and thus need a really compelling reason to be replaced.
SICP occupies an interesting place in this space: computer programs definitely had quite a bit of development, but Abelson and Sussman are amazing and their perspective remains valuable.
Haven't seen that computer architecture course, you seem to suggest it may be another of this kind?
The architecture course is 6.004, although it is a bit different than Abelson/Sussman. Abelson focuses heavily on software concepts while 6.004 has a strong hardware bent.
I've seen so many courses with newer course-dates but the lectures alone are from older editions of the same course.