Data Collection Broken for some nodes in testing

I upgraded barbrady to the latest testing on Thursday, and when I went to look at the bandwidth KSC report I noticed NaNs on most of the interfaces since the upgrade. One, eth0 on nen.opennms.org (node 24) shows the following in the logs:

2011-05-20 14:34:00,416 DEBUG [CollectdScheduler-50 Pool-fiber11] CollectorThresholdingSet: applyThresholds: Ignoring resource node[24].interfaceSnmp[eth0-000d566ffea8] because data collection is disabled for this resource.

So I looked at the webUI, and while the background color was green, the collect field said "Don't Collect" (see screenshot)

In the database:

opennms=# select * from ipinterface where nodeid=24;
id | nodeid | ipaddr | ifindex | iphostname | ismanaged | ipstatus | iplastcapsdpoll | issnmpprimary | snmpinterfaceid
--------------------------------------------------------------------------------------------------------+----------------
2128 | 24 | 216.216.217.254 | 2 | nen.opennms.org | M | 1 | 2011-05-20 14:32:35.504-04 | P | 2158
(1 row)

opennms=# select * from snmpinterface where nodeid=24;
id | nodeid | ipaddr | snmpipadentnetmask | snmpphysaddr | snmpifindex | snmpifdescr | snmpiftype | snmpifname | snmpifspeed | snmpifadminstatus | snmpifo
perstatus | snmpifalias | snmplastcapsdpoll | snmpcollect | snmppoll | snmplastsnmppoll
--------------------------------------------------------------------------------------------------------------------------------+-------
-----------------------------------------------------------------+-----------------
93062 | 24 | 127.0.0.1 | | | 1 | lo | 24 | lo | 10000000 | 1 |
1 | | 2011-05-20 14:32:35.504-04 | N | N |
2158 | 24 | 216.216.217.254 | 255.255.255.128 | 000d566ffea8 | 2 | eth0 | 6 | eth0 | 100000000 | 1 |
1 | | 2011-05-20 14:32:35.504-04 | N | N |
124475 | 24 | 0.0.0.0 | | | 4 | sit0 | 131 | sit0 | 0 | 2 |
2 | | 2011-05-20 14:32:35.504-04 | N | N |
125128 | 24 | 0.0.0.0 | | | 5 | sixxs | 131 | sixxs | 0 | 1 |
1 | | 2011-05-20 14:32:35.504-04 | N | N |
124474 | 24 | 0.0.0.0 | | | 3 | eth1 | 6 | eth1 | 0 | 2 |
2 | | 2011-05-20 14:32:35.504-04 | N | N |
(5 rows)

So it looks like a bunch of interfaces got deleted from the ipinterfaces table for some reason. And it is odd that the one interface I'm referencing still exists, and is set to primary, but that OpenNMS doesn't believe it should collect, but it should:

$ head /opt/opennms/etc/datacollection-config.xml
<?xml version="1.0"?>
<datacollection-config rrdRepository="/opt/opennms/share/rrd/snmp/">
<snmp-collection name="default" snmpStorageFlag="select">