We currently log progress every 100000 nodes, which can take a lot of time on slow systems.
- log based on a time interval, not number of nodes
- each time, log cluster node id being recovered as well
- log rate (nodes/per time unit)
not only log number of suspects but also nodes that need recovery