I didnt collect accounts at GM kiddo but by all means keep making things up, it sort of makes you sound more ridiculous.
As for what statistically significant sample size would be, well there is a LOT of math to do for that, which I honestly do not feel like doing right now, but if I remember right it was roughly 100 games of non placement ranked or about 1-2 serious seasons of playing (hence why an MMR reset breaks the ladder).
Now the other thing you had wrong was when you talked about having a Masters Zarya vs a Plat Zarya which is not the common occurrence. 1v1 does not matter. You are the constant. To simplify it, over time your opponents and teammates will even out so the only true constant is you. Basically if you are a 2500 player and you belong at 2900, you should hold about a 55-60% win rate. Now this means you could easily lose 4-5 in a row (though as you fall your chance to win goes up) because that is how numbers work. However over time you will push yourself back to your equilibrium point.
Furthermore, once you hit that equilibrium point it is actually fairly likely that you experience an 8 game win or 8 game lose streak to push you past it before you regress/climb to the mean, at least at some point. Once you hit the SR you belong at, whether you win or lose is effectively 50-50 and if, for example, I flip a coin 100 times I EXPECT to see 7-8 of either heads or tails in a row.
Basically MMR is a crutch people use to hold themselves back when if they simply looked at their own gameplay, or had other people do it as I do not mind doing it myself if you share a replay code, then you will learn and you will get better and you will climb, but the moment you fail to see that you are the only thing holding you back then you are done climbing and you are done getting better.