The only time I have seen anything similar was due to a bug in a particular network-card driver on either Dell or HP laptops (I don't remember which, but it was an Intel NIC) that flooded the network with a huge number of IGMPv6 join messages when the systems went to sleep. It happened about the same time every day because it was about 30 minutes after people went home and their PCs went to sleep. Anyways, we could tell there was an issue because of the high CPU on the switch at the time (had to be viewed by serial console as the VC in question was unreachable on the network), and eventually found the huge flood of join messages (>100k / sec). A driver update was the eventual fix, but in the meantime we created a firewall filter blocking IGMPv6 join messages and the problem went away.
Not sure if that is related to your issue, but thought I would point it out.
I would also check your spanning-tree topology to see if it is changing (show spanning-tree bridge) around the time of the issue. Having mixed-vendor layer-2 networks without a well thought-out STP topology can be very problematic in my experience.
-Ron