This is an extract from a GC log that was sent to us

1454.986: [CMS (concurrent mode failure): 9224K->10063K(843776K), 0.8197855 secs] 151890K->10063K(1014464K), [CMS Perm : 17494K->17270K(17536K)], 0.8207795 secs]
num #instances #bytes class name
--------------------------------------
  1: 38314 4811808 <no name>
  2: 60969 2784112 <symbolKlass>
  3: 38314 2762744 <methodKlass>
  4: 35377 1981112 java.nio.DirectByteBuffer
  5: 3480 1918464 <constantPoolKlass>
  6: 18353 1578184 [C
  ...

A SIGQUIT was sent to the JVM to get a heap histogram. Before that heap histogram could be calculated, a Full GC was done. The Full GC was flagged as "concurrent mode failure" which is incorrect as there's not even a GC cycle running at the time.