I find the 3333 part very puzzling indeed. It makes it look like a broadcast address, and not even just a multicast. As I am sure you know already, you take the first byte (33), and write it backwards in binary (11001100). The first bit says it is not station specific, i.e. broadcast or multicast. The second bit says it is broadcast rather than multicast. But in that case I would have expected it to contiunue as 0xFF. If these are in the CAM table, they may have been used as source addresses, which is illegal.
I don't know switch IOS too well, but in CatOS I would set the SYS logging level to at least 6. This would give me a syslog message every time I see a frame apparently from a multicast address (and I do see them in my network from time to time, but that's another story).
I would guess something is misbehaving in your network, sourcing frames from a bad address with the first two or three bytes corrupted.
I have a clue: it's a longshot but maybe ... 0100.5e00.000d is the multicast destination for PIM routers and 0100.5e00.0016 is the multicast destination for IGMP - both protocols concerned wityh controlling IP multicast. Have you introduced IP multicast into your network recently?
Try snooping the broadcasts on your network and see if you notice anything unusual. I have seen PCs generate frames from multicast addresses when they are in standby.
Kevin Dorrell
Luxembourg
