For example, if you have 9 workers in 3 groups and all workers in first group are dead (they exceeded the limit of sequential deaths), then an application will stop to serve requests on the dead group port. Master process don't restructure the groups on workers death and don't emit any errors or events about the issue.
Options:
- emit
group dead event and log or throw en error if listener is not assigned to event
- restructure workers groups until workers count is greater or equal to groups count and fail otherwise.