Comment 17 for bug 996369

Just though I would add a comment that I am also seeing this behaviour. It (might) be related to having a "post-up" command attached to the ethernet configuration. Here is my interfaces file:
xfers ~ # cat /etc/network/interfaces
# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto bond0
iface bond0 inet static
 address 172.16.1.10
 netmask 255.255.255.0
 broadcast 172.16.1.255
 network 172.16.1.0
 gateway 172.16.1.1
 dns-nameservers 10.0.0.120 10.0.1.120
 dns-search mps.lan wilts.mps.lan
 dns-domain mps.lan
 bond-slaves none
 bond_mode 802.3ad
 bond_miimon 40
 bond_lacp_rate 1
 bond_use_carrier 1
 post-up /usr/local/sbin/check-bond.sh $IFACE
 pre-down /usr/local/sbin/check-bond.sh stop $IFACE
 ## ftp.mps.lan - for internal access
 up ip addr add 172.16.1.20/24 dev $IFACE

 ## ftp-sohonet.mps.lan - for FTP/Aspera Connect over Sohonet
 up ip addr add 172.16.1.21/24 dev $IFACE
 ## ftp-abovenet-7a.mps.lan - for FTP/Aspera Connect over the Abovenet 7A link
 up ip addr add 172.16.1.22/24 dev $IFACE
 ## ftp-abovenet-7a.mps.lan - for FTP/Aspera Connect over the Abovenet 7B link
 up ip addr add 172.16.1.23/24 dev $IFACE

 ## faspex-sohonet.mps.lan - for FASPEX over Sohonet
 up ip addr add 172.16.1.24/24 dev $IFACE
 ## faspex-sohonet.mps.lan - for FASPEX over the Abovenet 7A link
 up ip addr add 172.16.1.25/24 dev $IFACE
 ## faspex-sohonet.mps.lan - for FASPEX over the Abovenet 7B link
 up ip addr add 172.16.1.26/24 dev $IFACE

# Slave Definition for bond0
auto eth2
iface eth2 inet manual
 bond-master bond0
 bond-primary eth2 eth3
auto eth3
iface eth3 inet manual
 bond-master bond0
 bond-primary eth2 eth3

If I comment out my post-up line, I appear to get a reliable network interface brought up at boot time, however, with the post-up line enabled, I often see one of the slaves fail on boot.