+1 for "500-1000" responses being a perfectly valid sample size. Basically every resource I've found says <1000 respondents gets you <5% margin of error. *Of course, you have to understand the implicit biases in your questions/available answers/and something about the response bias in order to use that knowledge.
Larian should definitely include a survey with the game, optimally one that pops up automatically the first time you close the game and then another one when you reach the end of the EA content. This would ensure that most players are reached.
+0 for "Too many overpowered encounters." This is either true or false depending on your definition of "overpowered." If you are going purely by D&D 5e rules/encounter calculator, then yes there are many BG3 encounters that would destroy a PnP party.
However, PCs are much more powerful in BG3: jump+disengage, high ground/backstab Advantage, BA shove, help restoring 1hp, everyone can use scrolls, and of course the unlimited resting. Particularly the last one: encounter balance in 5e is based off of the assumption that a party will face ~3-6 encounters in an Adventuring Day. A "Deadly" encounter becomes much easier if the party uses all their resources with the intention of long resting immediately after.
