This could be an oversubscription situation due to broadcasts (storm, L2 loop?) or
due to traffic flooding from several ingress ports directed
to lower-rate ports (Gig->FastEth for example).
Like some servers/hosts sudenly start putting excessive traffic on the network.
You should see a lots of output drops on the interfaces connected
to those oversubscribed ports as well, when it happening.
Messages looks like kind of indication of out-of-buffers conditions.
If it would happens again, in order to find from behind which ports flooding commes from
and which are oversubsrcibed make sense to check several time stats of
"show int", "show int switching" and possible "show spanning-tree detail".