I do think it's a book of quality, which admittedly is a relatively rare category. But yes, I think it's
hugely overrated, especially when characterized as the "one true best way" or something, to learn programming.
(Note that I am commenting on the first half or so of the book, never having gotten past that. Maybe the last half of the book is the best book ever written, I do not know.)
For one thing, half of the examples are calculus. If you are an engineering graduate student with experience in fortran (which is the original intended audience, just watch the original lectures), that's a great idea. For humans with little to no formal mathematical training, it's pretty poor. I once helped to lead a study group on SICP with half a dozen highly motivated junior-to-intermediate programmers, and everyone found the exercises insanely hard for this reason. (Of that group, btw, I would say one student greatly benefited, two dropped out, and the rest got some benefit.)
For second, I strongly disagree with the claim that functional is a "more natural" or "more intuitive" way to think about programming. I've helped teach hundreds of adult students to program (and become working developers), most of them with no prior experience, and I would say that less than 1% of them found recursion anything other than a crushing mindfuck. Whereas more than half of them took to imperative programming in a much shorter length of time (a few weeks, say). (None of these students were given SCIP, tho.)
Based on the perennial praise in threads like this, I concede that must be those in the world for whom it's a great book, and that's great. I am firmly convinced that these people are atypical.
On the other hand, it is a book that concerns itself with important questions, where IMO most books do not, so that alone makes it, as I said at the beginning, a member of a rare and distinguished category.