Yes, this is the conventional interpretation, and what will normally happen. But what @Kevin_Levites was talking about was the corner case where, by freak chance, all molecules have a position/velocity vector that is near/points towards the lid. Highly unlikely, yes, but still a valid microscopic state. It would, of course, quickly develop into a more likely state due to collisions etc. My gut feeling is that if you base your argument on the statistics of waiting long enough for this to happen, you should also wait long enough for other freak states to occur, in order to get valid sampling for the statistics and a proper time average. But in reality this whole scenario reminds me too much of a Maxwell demon-esque situation that I’m quite unwilling to accept it actually happening.