What I did is watching the console and set a debug print at the top of the trigger OB script side
I had the debugtriggers prints (all at once, thus making me think just use the one having a velocity, to know if the door is either opening or closing, since the button event is twice 'opening' on both events), then a delay before the debug print, that's what made me think there were a delay
I've followed the wiki page about path through switch, but it seems, as did Tomek, that moving the Enabled false from 'after the Opening/Opened' to 'before' helps forcing bots stop using the button, even though the condition in ExitConditions looks right (first instruction of trigger is to change door status from Closed (0) to Opening (1), and ExitConditions tests status > Closed, just like in the wiki)
IIRC most of OB is single-threaded, perhaps it stays somehow stuck in ExitConditions, and thus the status switch happens later; and as the map author did not protect the button from spamming, it gets the whole thing stuck; but why would setting Enabled to false would then make it react so quickly?
I'm not sure to understand the flow; perhaps the issue here is only that the button is unprotected by map author, contrary to some other maps?
Or am I totally misunderstanding things?