The Facebook "Graph API Rate Limiting" docs says that an error with code #4 is an app level rate limit, which is different than user level rate limits. Although it doesn't give any exact numbers, it describes their app level rate-limit as:
This rate limiting is applied globally at the app level. Ads api calls are excluded.
- Rate limiting happens real time on sliding window for past one hour.
- Stats is collected for number of calls and queries made, cpu time spent, memory used for each app.
- There is a limit for each resource multiplied by monthly active users of a given app.
- When the app uses more than its allowed resources the error is thrown.
- Error, Code: 4, Message: Application request limit reached
The docs also give recommendations for avoiding the rate limits. For app level limits, they are:
Recommendations:
- Verify the error code (4) to confirm the throttling type.
- Do not make burst of calls, spread out the calls throughout the day.
- Do smart fetching of data (important data, non duplicated data, etc).
- Real-time insights, make sure API calls are structured in a way that you can read insights for as many as Page posts as possible, with minimum number of requests.
- Don't fetch users feed twice (in the case that two App users have a specific friend in common)
- Don't fetch all user's friends feed in a row if the number of friends is more than 250. Separate the fetches over different days. As an option, fetch first the app user's news feed (me/home) in order to detect which friends are more important to the App user. Then, fetch those friends feeds first.
- Consider to limit/filter the requests by using the following parameters: "since", "until", "limit"
- For page related calls use realtime updates to subscribe to changes in data.
- Field expansion allows ton "join" multiple graph queries into a single call.
- Etags to check if the data querying has changed since the last check.
- For page management developers who does not have massive user base, have the admins of the page to accept the app to increase the number of users.
Finally, the docs give the following informational tips:
- Batching calls will not reduce the number of api calls.
- Making parallel calls will not reduce the number of api calls.