Graphs for one device suddenly stopped updating

[uno]

Hi.
We are graphing some 130 devices in our cacti. Yesterday morning, all graphs except two for a device suddenly stopped updating. The device is a Cisco 6509 switch.

The cacti log shows lots of messages like this, one for each non-working datasource:

01/31/2013 10:45:02 AM - SPINE: Poller[0] Host[191] TH[1] DS[5651] SNMP: v2: r2.domain, dsname: traffic_out, oid: .1.3.6.1.2.1.31.1.1.1.10.40, value: U
01/31/2013 10:45:02 AM - SPINE: Poller[0] Host[191] TH[1] DS[5651] WARNING: SNMP timeout detected [500 ms], ignoring host 'r2.domain'

I have troubleshooted SNMP on the switch, and it works perfectly fine. I also added a secondary device to cacti, graphing the same switch and all those graphs work.
I tried replacing the rrd file on one of the old datasources with one from the test device, and vice versa, and then the old graph (with the new rrd file) works and the new graph (with the old rrd) doesn't.

Does anyone have any idea what the problem might be?

[gandalf]

Please gather some more data: see 2nd link of my sig
R.