chunked/nltk_chunking/_pipeline_webserver/chunk_18.txt

Next, the timeout is reset on every queue fetch, meaning you could
wait much more than 1ms before running the inference (delaying the first request 
by that much).