Make failure explicit

External services fail, requests get repeated and background jobs stop midway. I prefer to account for those cases as part of the design instead of adding fixes after production incidents.