Comment 64 for bug 274405

Still a problem in Maverick. Takes over thirty seconds to reconnect after resume from suspend.

dmesg | tail
[ 861.369996] sky2 0000:02:00.0: eth0: phy I/O error
[ 861.370083] sky2 0000:02:00.0: eth0: phy I/O error
[ 861.370171] sky2 0000:02:00.0: eth0: phy I/O error
[ 861.370258] sky2 0000:02:00.0: eth0: phy I/O error
[ 861.370345] sky2 0000:02:00.0: eth0: phy I/O error
[ 861.374828] sky2 0000:02:00.0: eth0: enabling interface
[ 861.376178] ADDRCONF(NETDEV_UP): eth0: link is not ready
[ 864.147940] EXT4-fs (sda1): re-mounted. Opts: errors=remount-ro,commit=0
[ 865.152547] EXT4-fs (sda1): re-mounted. Opts: errors=remount-ro,commit=0
[ 872.192056] eth1: no IPv6 routers present

Also, still seeing this in comments:

cat /usr/lib/pm-utils/sleep.d/55NetworkManager
#!/bin/sh
# If we are running NetworkManager, tell it we are going to sleep.
# TODO: Make NetworkManager smarter about how to handle sleep/resume
# If we are asleep for less time than it takes for TCP to reset a
# connection, and we are assigned the same IP on resume, we should
# not break established connections. Apple can do this, and it is
# rather nifty.

It is nifty. It's been nifty for a while. Any chance it'll get implemented for Natty?