Drastically reduces the size and sparsity of the quinn receive buffer for no performance cost in typical applications.
Fixes #1608. Alternative to #1611 with less impact on endpoints that are heavily loaded by many peers. Not certain which we should prefer yet; recvmmsg might not be buying us all that much, but neither is supporting jumbo frames (outside of loopback). Maybe we should even do both!
Drastically reduces the size and sparsity of the quinn receive buffer for no performance cost in typical applications.
Fixes #1608. Alternative to #1611 with less impact on endpoints that are heavily loaded by many peers. Not certain which we should prefer yet;
recvmmsg
might not be buying us all that much, but neither is supporting jumbo frames (outside of loopback). Maybe we should even do both!