You agree that OpenAI is choosing the voice because it sounds like SJ. How exactly is that different from impersonation?
You don't know that's what happened, but it wouldn't matter either way. Regardless: it is misleading to call that person an "impersonator". I'm confident they don't wake up the morning and think to themselves "I'm performing SJ" when they order their latte.
The key here is intent. If there was no intention for OpenAI to model the voice after the character Samantha, then you're right, there's no foul.
But as I have explained to you elsewhere, that beggars belief.
We will see the truth when the internal emails come out.