Not my lucky day.
Eventually I managed to run that server on my machine, but... but the server crashes pretty frequently, with obscure exceptions from aws. Well, at least I could add "println" to see that it receives the requests. Ok, it receives GET, but not POST. Maybe it does receive POST, but I don't see any trace of that, and curl does not get any response.
Ok, and of course I tried to do POST on staging, via curl or via swagger; it returns ok there (but not locally, the same request) - but it does not append the data I send.
So, fck, it's a waste of time, but... but I have to do it.
Also, I tried to reproduce the behavior on production, and kaboom, I can't log in (as a regular user), and more, it says I have a wrong password, and when I try to "reset the password", it sends me an email with a wrong url, and when I ask my colleagues where t.f. should I file frontend bugs, nobody knows.
OTOH, at 12:30 had a pretty good meeting with our partners (with the presence of Karen and Renato). It turned out, Bhargav is a pretty bright guy, and he did implement all that push absolutely correctly, so all my caveats are perfectly covered. He also managed to explain to me why the data look so weird. It's not a duplication, they pass the correlation, and it's on our part that the implementation sucked.
I'm less and less impressed by those nerds that slapped together the code, not understanding the meaning of data (I bet nobody knows what "cume" is), missing tests, not providing solutions for proper testing and debugging. Хуяк-хуяк и в продакшен.
Ok, what can we do. As much as we can, right?