Except he didn't really "resign" Linux to x86 in any substantive way. Ted Ts'o points out later in that thread that Linux, even at that early stage, was really not that intimately tied to x86; Minix certainly has architecture-specific code as well, and there was never really an objective comparison of how much of each system was dedicated to a particular machine architecture.
One of the points of contention where Linus was emphasizing 386-specific code was really not about portability to other
architectures, but rather portability to earlier x86 processors before the 386. At that time, this was somewhat of a bigger deal: it wasn't backward compatible to pre-386 x86 processors because it used 32-bit protected mode. Tanenbaum criticized him for taking special advantage of these 386-specific features -- he said, " I don't have figures, but my guess is that the fraction of the 60 million existing PCs that are 386/486 machines as opposed to 8088/286/680x0 etc is small." Linus's defense of 386-specific features wasn't really about "resigning" Linux to x86 per se, but about using 32-bit protected mode with modern MMU features which weren't available on earlier processor models. Sure, other architectures weren't on Linus's radar at the time since it was a small project, but I don't think he ever claimed he'd object to the premise (not that I'd put it past him).