This project runs on 2 fairly small servers and here's a story on how I investigated and fixed this CPU load issue that was bogging me for a few weeks.
It actually uses screen scrapping as primary resource and the api just to read stuff out. It's the exact same request and interaction as a real person would do with a browser, so no violation of ToS.
Yes it is. You don't get banned because the robot works inside Instagram's API daily limits. It's just like a regular person doing likes here and there during the day.