It may be observed hundreds a times a day, it may just be that number of OSM mappers in general population is low enough that no OSM mapper ever observes a function…
But, yeah, Verifiability - OpenStreetMap Wiki is a thing. It is easy to verify something has a button. It is other thing altogether (and hugely harder to verify!) what that button actually does.
Even more, many mappers will not invest enough effort to investigate such using proper scientific method. E.g. someone using StreetComplete might come at the crossing, see 4 boxes with button on each, and confirm it is button_operated=yes before following along to next quest few seconds later.
Properly doing statistically significant number of measurements for each crossing, spreading along all imaginable influential factors (time of day, time of year, weather, traffic, different button presses patterns - at different times of traffic cycle, once or several times etc.) and tracking differences along those, and then trying to prove correlation versus causation, is worthy of an academic dissertation; and not for more precisely mapping low-utility OSM tag.
But yeah, if one is pretty certain that it does nothing even after prolonged observations, it might be mapped with button_operated=placebo and appropriate note=* (otherwise, if button_operated=* is just missing, very next mapper will add it, and thus endless cycle of useless adding/removing tag will begin 
)