instalation breaks network interface settings

Actually this is not quite true. It turns out on the next machine that this change was done earlier (Friday 13th, with interfaces.bak-0 file having the uncommented line), but it results in the system breaking once rpcbind, nis, etc has been added.

So before rpcbind & nis is installed, I can reboot and it works fine with the #NetworkManager# comment in place, but once installed it no longer brings the interface up unless I edit the file to remove #NetworkManager#

It is strange. I have a script to install NIS and autofs, it installs the following packages:

portmap
nis
tcsh
nscd
autofs
rsh-client
rsh-server
nfs-kernel-server

Then it creates/updates some of the file (auto.master and so on) to suit out use.

If I install each package in turn and reboot, it comes up (but with around extra 30s delay or so later once NIS is installed). But if I run the script to install all in the one go then reboot I find that eth0 is out of use!

Removing the comment appears to make it boot to ping/SSH access quicker.

I guess this bug is probably a network manager one rather than rpcbind

I think this is unlikely to be a bug in either rpcbind or network-manager, but a bug in how your script is configuring things. It's normal for network-manager to comment out dhcp-only interfaces in /etc/network/interfaces, and for these network devices to be managed by network-manager; and rpcbind should in no way interfere with NM bringing up interfaces. This all works out of the box here for me and has for several releases.

The odd thing is I ran the script after installing the packages one-by-one and it worked OK. Obviously it realised the packages were already installed so skipped doing anything there, then it set up the various files. However, it was still a bit slower to come up after reboot than when the comment-out section was removed.
The implication being something odd about installing several packages in the one apt-get command, which seems just as odd, or that the network manager had updated something in between (on the reboot between package installations) that prevented the loss of eth0.
Still, we need to change this entry anyway to a fixed IP address so it will become unimportant. Shame the reason is not known though.

Status changed to 'Confirmed' because the bug affects multiple users.

Well, I don't know why someone else has just marked themselves as affected by this bug, but the bug is unreproducible and we have no leads on a solution; so I'm closing it as invalid.