On commit fdc8828e8527b6fa4997072fb4795e4060c68cc0 (tip of master as of the time of this comment), I am unable to reproduce the behavior. I experimented with a variety of curl timeouts (the -m argument), including the proposed 0.001, but still could not reproduce this.
On a freshly-restarted VM, I did see a small jump in the number of non-TIME_WAIT connections after the first request, but that jump did not occur on the second and subsequent requests, so it appears to be something warming up.
I also checked stable/liberty (47eb6a37) and was unable to reproduce the behavior there either.
I *can* reproduce this on stable/kilo (da428c4f).
I suspect that commit 12d8a53 fixed this, though I have no proof of that.
On commit fdc8828e8527b6f a4997072fb4795e 4060c68cc0 (tip of master as of the time of this comment), I am unable to reproduce the behavior. I experimented with a variety of curl timeouts (the -m argument), including the proposed 0.001, but still could not reproduce this.
On a freshly-restarted VM, I did see a small jump in the number of non-TIME_WAIT connections after the first request, but that jump did not occur on the second and subsequent requests, so it appears to be something warming up.
I also checked stable/liberty (47eb6a37) and was unable to reproduce the behavior there either.
I *can* reproduce this on stable/kilo (da428c4f).
I suspect that commit 12d8a53 fixed this, though I have no proof of that.