I would be very surprised if it was a loop. We don't have issues anywhere else that would support it.
This all happened when we changed our primary profile. Before we had one /24 network for phones and PCs (the PCs connect through the switchport on the phone) and the E1 port on the RAPs (LAN side) was trusted and no user roles were defined. I am trying to get confirmation that it was working before on this profile.
We ran out of IPs and divided the pool into 2 new /24 networks, 1 for phone and 1 for PCs. It all has to run on DHCP so we had to do some work. We trunked E1, untrusted it and set the native VLAN to 1000 (the PC VLAN). We enabled option 191 on the DHCP server which uses fingerprinting to ID the phone and tells it to initiate a new request in VLAN 1001 (phone VLAN). Also we set the default rule to authenticated and used a MAC based UDR to identify the phones and put them in the voice role. We were working fine for about a week with ~20 users on the profile. One night we moved over about 100 users and the circuit was pegged. We enabled BCMC on VLAN 1000 and the usage went backed to normal.
We use about 20MB of the 100MB circuit for normal operation and under 5Mb for off hours, we were up over 75MB during peak utilization and around 20MB for non-peak once the new profile was in place.