As the Execution Model mentioned,
By default, Logstash uses in-memory bounded queues between pipeline stages (input → filter and filter → output) to buffer events.
but there is another saying in Memory queue,
By default, Logstash uses in-memory bounded queues between pipeline stages (inputs → pipeline workers) to buffer events.
And I've seen something that logstash optimized the model, combine filter and output to one thread, So is the statement in Execution Model is wrong? there is only one queue between input and pipeline workers?