Both are good questions. With regards to the RCP average, multiple forecasters with good reputations say that the issue with the RCP average is that they don't account for "house-effects", i.e. the right-ward or left-ward leaning bias that a poll may have. I think the 538 forecast (
https://projects.fivethirtyeight.com/2020-election-forecast/florida/) is generally pretty solid as they account for this sort of bias while remaining conservative in their estimates and allowing for more uncertainty in the data. 538 do include pollsters like Trafalgar and Zogby so that they are at least accounting a little bit for any truth that could be present in that data. At this moment in time, their forecast will predict a tighter race than the polls may suggest (1.4% in FL as of today) and then, if the polls maintain the same trend closer to election day, the forecast difference between Biden and Trump will widen. I think the other thing to look at is that in most polls, Biden is close to or at 50%, which is a good sign that any undecided voters would likely not change the final result, unlike 2016. Margin of victory in the polls isn't as important in this scenario. Of course, effective voter suppression can throw all of this out of the window as forecasts and polling only look at "intended voting", not whether votes will be cast and accepted or thrown out.
I am not completely sure how Florida will be faster to call on election night than other states but have seen multiple reputable sources recently mention that FL has better vote reporting mechanisms than most other states, which could allow for a victor to be credibly declared on election night. Again, not sure how that works if true.
EDIT: Should also mention Rasmussen as a pollster to watch out for in having significant right-leaning bias. They were completely off in the 2018 midterms as a result. Still it's good to include these polls and adjust them in any final analysis to cover all bases.