The collection chunks defragmentation algorithm info involves the submission of moveChunk commands from the ConfigServer through BalancerCommandsScheduler.

The involved shards may require tens of seconds to complete such operation, during which an election of a new ConfigServer primary may occur: the new primary will have to resume the defragmentation algorithm from a stable state - meaning that no new commands may be issued as long as all the moveChunk operations launched by its predecessor have completed.

Such logic needs to be implemented within BalancerCommandsScheduler relying on the persisted information about previously issued moveChunk operations (see SERVER-59964 for details).