If one of the indexer in the indexer cluster has blocked queues, which one of the following scenario is possible.
1) listen to the incoming events and Drop events that are received by it.
2) Stop listening to incoming events. So that data is forwarded to other active indexers in the cluster.
Generally speaking, the reason to have multiple Indexers is to address a scenario where one Indexer (for some reason) isn't accepting connections. If the queues are blocked, the Indexer will stop accepting connections. The Forwarder's will cycle though to the next Indexer in the list, as defined in outputs.conf. That's your #2. The only time I've run into your #1 scenario is if a Forwarder or Indexer is forwarding data to a 3rd party, that 3rd party system stops accepting connections, and the outputs.conf settings such as dropEventsOnQueueFull
are configured to allow a drop.