Traffic apps only know about congestion if someone running the app goes down the congested road. Because of this, I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.

Doesn't have to be true. My state has public camera feeds for various highway stretches. Too low-detail to identify vehicles but easily enough to detect congestion.

>I've always suspected that the apps, from time to time, will route someone down a route they haven't gathered data on in a while, just to collect the data, and even if the route is likely to be suboptimal.

I can't say this actually happened for me. For straightforward routes with no congestion I never saw random alternate routes being proposed. That makes sense, given that it'd probably tip people off. If this is happening, they must only be doing in cases where there's congestion and the difference is marginal, eg. it's rush hour and the "optimal" route takes 30 minutes but the alternate takes 33 minutes. Moreover you don't really need any deliberate effort to see this effect. If nobody is traveling on a side road, the algorithm will probably revert to historical patterns, which might turn out to be overly optimistic in congestion scenarios (eg. there's nearby road repairs and other people are already using it as an alternate), thereby giving you the impression that you got screwed over by the app.

I wonder this same thing when I’ve been routed completely asinine ways off interstates at night when traffic appears calm. I always have to play the game of “is there horrendous traffic ahead from night paving / accident, or am I being conscripted as a traffic sensor robot?”