It will depend on the number of clients making requests and what percentage of their time is spent making requests.
Imagine a client that takes a second to upload all the data for a request. If it does that every other second it will use up half a worker thread. If it does it every second it will need a dedicated worker thread. If it does it once a minute a single thread might be able to handle dozens of clients.
Then there is the question of how patient the clients are. If every client makes an http call at the top of the minute and only waits a few seconds before declaring the logstash server to be non-responsive then you need enough threads to do all the work in those seconds, and they can sit idle until the next minute. There are many variables.