A possible confounding factor which leaps to mind is the notion of register: almost all language instruction is in a utilitarian register, and daily L2 use is also commonly utilitarian; indeed, emotional surprise tends to cause people to fall back* on their L1.

If we could find people who learned an L2 in an emotionally charged environment (at an orphanage? in a teenage gang?) but never subsequently used it for work or leisure in a utilitarian environment, would they show the opposite results?

* I've heard that in WWII, a standard counterintelligence ploy was to suddenly stomp on an interviewee's foot to observe if they reacted with "Ow!" or "Aua!" or "Aie!", etc. (this reaction is trainable, of course, as the number of french who say "punaise" [minced, like "gosh darn"] when they hit their thumbs with a hammer will attest)