Calls that are longer than the check interval never get returned. The beginning and end of a call both have to be within the check period to get counted. Increasing the check interval is a temporary mitigation that introduces some more latency and isn't flexible.
Keep track of existing encrypteds in a TTLCache (cachetools package is super good for this) or a deque (I think that exists in the collections stdlib package?) and check that cache on every API call, skip calls that already exist and add the new ones
Tweet only the new calls
Add some mechanism so the bot doesn't blurt out a bunch on startup if there are a number of recent calls
The OpenMHZ web app uses Socket.io to push updates to clients, maybe the bot could get a session going and get call data through that instead of API scraping.
Calls that are longer than the check interval never get returned. The beginning and end of a call both have to be within the check period to get counted. Increasing the check interval is a temporary mitigation that introduces some more latency and isn't flexible.