Ancalagon
Dusty Dragon
Good point. I'd extend that to say that you need to trust your table. It's two separate, but related, issues.
The first is the spotlight issue. If the players at the table have conflicting ideas about what is "fun," that can cause problems. It is very difficult for even the best DM to adequately design campaigns and encounters when one (or a small number) of players derives their fun from wringing all possible mechanical advantage, while the rest of the players do not ... and vice versa. Or, put another way, it tends to work best when the players have a general consensus as to the style they are going to play. Otherwise you end up with the old "Angel Summoner and BMX Bandit" problem, which tends to lead to table strife.
The second is what I call the treadmill. The reason that (IMO) many tables don't worry overly much about efficiency after some period of time is because difficulty is just a treadmill of increasing speed. If everyone is optimizing, then the DM has to increase the challenge levels. Absolute difficulty might be increased, but relative difficulty remains the same. And all that really matters from the player perspective is relative difficulty. In my opinion, given the constraints of most D&D editions (including 5e) you can maintain a sweet spot of play for far longer if you don't try and wring every possible mechanical advantage. YMMV.
This "not everyone optimizes" is a big problem in 3.X and pathfinder 1e games. The gap between an well optimized character and non-optimized one is immense, esp when you start hitting level 7 or so (your angel summoner/BMX bandit comment is on the nose). Reducing that gap to a much more manageable level was a significant improvement in 5e. You can optimize if you want, and you will be better, but only a moderate bit.