We do not expect increasing the number of partitions would improve migrations performance. For large data sets, some operations may be marginally improved by having more partitions.
We have tested internally with cluster > 100 nodes but there is a hard limit of 127 node in an Aerospike cluster. The largest production cluster that I am aware of was about 48 nodes. I wouldn’t expect the number of heartbeats to be a problem, but if it were they can be tuned down.
In the 3.5.x releases, migration has been significantly improved. But if a server were to be stopped and then started all of the partitions that were assigned to it would need to be resynced which will take some time. With more nodes in the cluster there will be less partitions assigned to any particular node which results in less partitions migrated for single node outages.