BBR is a moving target

[mattmathis]

Placeholder issue that we need to track BBR versions and consider loading tcp_bbr that is newer than the CoreOS base kernel.

Re the BBR behavior mentioned in ndt-server/#132

Neal Cardwell says:

BBRv1 does have the property that it will only exit "STARTUP" mode once it has had 3 non-application-limited bandwidth samples that fail to push up the estimated bw by 25%. This means that application pauses (within a single round trip time), which "invalidate" bw samples by marking them as application-limited, prevent BBRv1 from exiting STARTUP.

This is not fixed by the January 2019 BBR patches. It should be fixed in BBRv2.

[pboothe]

Speaking of BBRv2 - https://github.com/google/bbr/blob/v2alpha/net/ipv4/tcp_bbr2.c