What’s required for that isn’t “model parameters”. It’s the resemblance between the observed sequences and the sequence expected of an ERV.
Good question. Have you looked for data? I havenm’t.
I would suggest that the answer to both questions would be “rarely”.
Still, the answer would seem obvious, given that most ERV insertions are in junk DNA.
What’s the probability that a designer would put the same sequence in separately created species, and what’s the probability that the sequence would look just like an ERV?