It probably depends if he's one of so many ignorant "grammar nazis" that don't understand how language works. Once you wade through the thin veneer of rule vomit, most prescriptivists have little to no clue about the language they are enforcing -- only demonstrating an ability to remember lists of uninformed, anachronistic and often ignorant rule sets that are usually intended for application in a specific milieu. They're usually foolish enough to then go and demonstrate this nonsensical approach in public. Usually as a pedantic and insufferably pretentious, high-friction social interaction: correcting minor typos, misapplying and enforcing meaningless rules, "correcting" conversational and idiomatic prose into stilted formal styles (thereby losing nuance and ultimately meaning). They then claim that it's important they do this because language is used to communicate, but they then demonstrate an almost perfect ability to use language in a non-communicative fashion.
On the other hand, even descriptivists can get a little frustrated when "too" and "to" are mixed up; "their", "there" and "they're" are conflated; "its" and "it's" are misused, and so on. It does demonstrate somebody who is unfamiliar with handling the language in the written form, but indeed the correlation between intelligence and understanding the complex English grammar in its fullness are undeniably low. It may simply be that a programmer with poor grammar has spent more time handling code (which also happens to be a milieu with very clear right and wrong feedback mechanisms for incorrect syntax and structure) than in handling written language. Feynman for example, wrote very little prose of his own.
tl;dr This employment tests only filters for people who have spent a great deal of time handling language and provides almost no information on their time handling code.