So we can know when we're getting close and can start thinking about forking to a child workflow (like Temporal ContinueAsNew).

Right now we have no idea when we'll hit a wall if there's potentially arbitrary looping (e.g. waiting for an export from an external platform to finish)