User blog comment:Koteitan/Purely mathematical definition of BMS/@comment-24923514-20191106231524/@comment-24923514-20191107231444

> In that case, P is just undefined.

What happens if you're evaluating expand(S[n]) and, because certain ancestors don't exist, the bad root doesn't exist? Or does the bad root always exist for the matrices that we care about?

And thanks for the permission! I won't change any of the information, I'm just going to rephrase some things to make it easier to understand.