Each task here is a request an analyst gets every few weeks, written in business language. Translate it into a pattern first: is it a duplicate problem, a streak, a session, a funnel, a cohort, or a period comparison? Then write it in steps with CTEs, checking each step's rows before adding the next.
The data is a mix of tables you know. The edge cases from the unit apply: clean keys before deduplicating, count people rather than events, and fill missing periods before comparing them.