From version 2020.3 onwards, there is no longer the concept of “Leader” and “Follower” nodes; now nothing relies on there being a single leader.
This is due to a change to store the next run time for a job in the database rather than in memory. Any node can then pick up that configuration and determine whether the job needs to be run. The load, run, and save of the configuration is done under a cluster-wide lock, which prevents jobs from being run on multiple nodes at a time.