Honestly, I think that the first option that was proposed in the video would be counter intuitive to what you are trying to establish. If you want new players to stay, but punish on a single infraction, then odds are, they won't stay.
Plus, how would we determine that single infraction? What would be the parameters? If they were mildly toxic, would the get the immediate ban? What if the only thing they said the entire game was "noob." Permaban?
I think that system would be a bit too harsh.
The second system is no better, IMO. If someone says the aforementioned "noob" and gets a warning about being permabanned, would they really want to stay? In a system that is so harsh? It would alienate new players as much as a toxic player might. I mean, having a toxic ban system is probably worse than having a toxic person, that you can just mute.
I think the biggest flaw in your argument is the fact that you are using your own experience in this video. You have no other evidence other than that of your own, you have no "new player experience," because you aren't a "new player." The system in place determines whether you're a smurf or not and decidedly places you with other smurfs, which should be the case. Your skill level is typically determined by ten games, if I'm not mistaken (I feel like I read this somewhere, although I wouldn't be able to produce those links as I don't remember where I read it. So, take it with a grain of salt.)
Your own experience should be coupled with evidence from several other experiences to provide more accurate results. This is called research. Rather than make assumptions based on a single experience, research is required to make an educated theory.
I'm not saying that there isn't an issue. Everyone knows that the system isn't perfect, but I don't think your suggestions would really fix anything. In fact, I feel they would do more harm than good.