That said Florida does have a large number of electoral college votes and can swing, so it makes sense as a a choice of the designer.
Or think another way. The largest number of states that can be blue and red still wins is 4. There isn't any combination of any combinations where 5 of selected states are blue and red can still win.
To calculate actual probability don't look at just the number of read leaves to number of blue leaves. You need to take into account probability of each of the state being blue or red. I am too lazy to calculate it but at first sight red outcome seems extremely unlikely.
If Biden wins Florida, then it's a >99% chance of a Biden victory.
But if Trump wins Florida, it's a 58% chance of a Biden victory.
So Florida is essential for Trump, but not for Biden. Biden has a 70% chance in Florida.
So I wouldn't say the fate of the world rests in the hands of Floridians. However, an early Biden win in Florida could allow networks to call the election on election night rather than having it drag out for weeks like it otherwise might.
The 538 model factors in the outcome of Flordida's election with its own belief about the other states results. This is why if you select Biden wins Texas, the model believes the national result likely in landslide territory, the Biden expected electoral college outcome goes up +70, even though Texas only has 38 electors. If Biden wins Texas, he also probably wins in Georgia or maybe even Arizona.