Question: Does tensorstore avoid multiple in-flight requests to the same data?

Say I have a remote kvstore (s3/gcs/http) with caching configured and issue two requests for the same data in quick succession before a response is received: Will tensorstore link the second future to the first? i.e. will it avoid issuing the second request instead linking the first's result to the second's future? Or will it issue two simultaneous requests for the same data?