Compute churn with re-subscriptions

You are analyzing a free-month subscription experiment. Assume experiment eligibility has already been computed; every user in the experiment table is already qualified. Use the following tables: **Table: `experiment_users`** - `user_id` STRING — unique user identifier - `assignment_ts` TIMESTAMP — time the user was assigned to treatment or control - `treatment_flag` STRING — either `treatment` or `control` **Table: `subscription_events`** - `user_id` STRING — user identifier - `event_ts` TIMESTAMP — event timestamp in UTC - `event_type` STRING — one of `start_free_month`, `first_paid_month`, `renewal`, `cancel`, `reactivate` - `monthly_price` NUMERIC — monthly subscription price recorded on paid events Assumptions: - Each user appears once in `experiment_users`. - `event_ts` is unique within each user. - Do **not** re-implement qualification logic; assume `experiment_users` already contains only qualified users. - A user is **paid_converted_30d** if they have a `first_paid_month` event within 30 days after `assignment_ts`. - A user is **retained_paid_60d** if, at exactly 60 days after `assignment_ts`, their latest subscription state at or before that timestamp is a paid state. Treat `first_paid_month`, `renewal`, and `reactivate` as paid states; treat `cancel` as not active; `start_free_month` alone does not count as paid. - Users may cancel and reactivate multiple times. The latest state on or before day 60 determines whether they are retained. - All timestamps should be interpreted in UTC. Write a SQL query that returns the following output columns for each `treatment_flag`: - `treatment_flag` - `assigned_users` - `paid_converted_users_30d` - `retained_paid_users_60d`