Maybe Parler was deciding based on HTTP status codes returned from the Twilio API, or the presence of some expected "error" field in the response (if it's not there, consider it ok), and one side screwed up the handling of an unexpected case? Pure speculation.

Instead of having "if explicitly ok, proceed", they maybe had "unless it's an explicit negative response in an expected format (or a connectivity issue, timeout, etc.), proceed"?

Which would be a pretty dumb thing to do, but I've seen worse.