This is a little phenomenon that is not common but not unknown. I have guys test for this in class because if you have it, there’s more to it that can go wrong than what you have experienced.
First let me coin a phrase. It might not be new, it might be known by some other term elsewhere / previously, but this is what I call it: overcock. It simply refers to the fact that as any hammer-fired system cycles the bolt carrier (or slide) back to rotate the hammer into the cocked position, naturally it rotates it a little further back than the final cocked position. In the case of the AR, it has to do this for the hammer to be caught by the disconnector which holds the hammer back until we release the trigger, at which time the hammer drops back onto the primary sear, if all is right (sometimes that doesn’t happen, a whole 'nother, and more common malady, resulting in “fire on release”.
Overcock.
If you look at your AR with the upper off, with the hammer in the cocked position, and then rotate the hammer back a little to simulate a bit of overcock, you will notice that the trigger moves too. That’s normal.
With the selector on safe, if you pull the trigger, you will normally get a little bit of trigger movement before the trigger comes to a hard stop on the selector shaft. That’s normal too.
Your safety is probably “bank vault tight”. In other words, with the hammer cocked and the selector on safe, you get zero trigger movement. Well and good until the hammer is held in overcock by a bolt locked back. Now things are beyond tight and into interferance-land. The hammer is held in overcock by the locked-back bolt, which is pressing the front of the trigger down, which is trying to rock the rear of the trigger up, but your bank-vault-tight fit between trigger and selector are preventing the movement.
Often enough it will put enough bind on things that cycling the bolt with the selector on safe will impede bolt carrier movement.
The solution is to remove a little metal from either the trigger or the selector where they meet when the selector is on safe. I don’t see where it makes much dif which one you modify.
Why does it happen? That old term “tolerance stackup” applies here-- you have so many things influencing this relationship that I reckon tolerance stackup could explain it-- or, who knows, one or more of the involved dimensions is actually out of tolerance.
I see this in approximately 1 in 200 AR’s. Can’t say that I’ve tracked which brands if any suffer from it more, but “fire on release”, again something much more common, definitely occurs more or exclusively in the lower-end brands.
Jeez, wouldn’t ya think people would check for this stuff before shipping a gun!?