Seems to be another regression since
JDK-8239786: the global data would be added from cycle data, even if cycle data counters are zero. It was masked by global data printout filtering when counters are zero, but it is not actually correct: once at least one event happens (e.g. stray degen gc), we would see lots of zero samples. It would appear degen gc had been running with 0ns on every cycle.