The Morris worm affected around 2000 VAX machines a couple of years previously, and was the first ever such incident on that scale. In other words, almost nobody in 1990 had been affected by a computer security incident. It didn't make sense in 1990 to prioritise this security threat over efficiency concerns.
Insisting on memory safety back then would be like insisting on code being accompanied by checkable formal proofs of correctness now: It's a technique that can be applied right now and that does improve safety, but it comes at such a cost that the tradeoff only makes sense for a handful of niche applications (aerospace, automotive, medical devices).
Insisting on memory safety back then would be like insisting on code being accompanied by checkable formal proofs of correctness now: It's a technique that can be applied right now and that does improve safety, but it comes at such a cost that the tradeoff only makes sense for a handful of niche applications (aerospace, automotive, medical devices).