Java driver speculative requests: exactly 90 seconds read request wait time if speculative requests enabled.

Good time of day everyone, we've got the following thing happening.

4 node cluster, apache Cassandra 3.11.1

While speculative requests are enabled on the client's side (java datastax driver, maxSpeculativeExecutions=1 due to small cluster size), no matter which constantDelayMillis is set, driver waits exactly 90 seconds before sending request and getting responce from speculative requests.

Question: where these 90 seconds might be coming from?

"90 seconds" timeout or pause is not present anywhere in the client app's code or our Cassandra server-side timeouts.

Honestly, I'm a bit out of ideas where to look. Where the problem might be? Even "lame"/"easy" hypotheses are welcome, because my vision actually might be blurred with the search.

Thank you all very much in advance.

Best regards, 

Vsevolod Filaretov.