I don't think any of his code examples are really JTBs because, like you said, its not "true". In the cow example, you make an assumption based on the papier mache cow that actually turns out to be true through information you didnt have access to. All of his code examples are more about him making assumptions that turn out to be false instead based on information he didn't have access to (or rather didn't think to access at least).

I don't think this really affects the take home message from the piece, I'm just being pedantic that it doesn't parallel perfectly (which he even acknowledges people may say in the last paragraph).

He simply wasn't speaking with precision. If you replace "code change" with "pull request" in his statement, it's JTB.

No you're talking about the first example. On the second example he says

> But—gettier!—the email service that the code relied on had itself gone down, at almost the exact same time that the change was released.

So the error was caused by the email service going down, which is completely independent of the code change/pull request.

That is correct - it is then a JTB - but not a “Gettier”. None of his examples are Gettiers.