With null link flickering test, scaling onos from 5 to 6(7) nodes causes all nodes thrashing and very low topo performance

1) Run Baremetals cluster with null link flickering rate set to 1000 events/s on each node;
2) scaling from 1 ~5 nodes, topology event performance is "normal": 1 node: ~800e/s; 2nodes: ~1400e/s; 3 nodes: 2200e/s; ..... 5nodes: 3500e/s;
3) when scale from 5 to 6 (or 7) nodes, each node starts to thrash - high cpu util; event performance drop to single digit.
4) upon shutting down cluster to have 3~4 nodes remaining; topo event performance regain to the earlier level.