Answers whether this node is the one that should arm timers.

AshWorkflow.Scheduler.Precise holds its deadlines in memory. Two nodes holding the same timer fire the same deadline twice, so exactly one node may arm them. This behaviour is how Precise asks which node that is, and it is one callback so that a deployment can answer the question with whatever it already runs.

The implementations

• AshWorkflow.Scheduler.Leader.Single — always the leader. The default, correct on one node, and what every test wants.
• AshWorkflow.Scheduler.Leader.Oban — delegates to Oban.Peer.leader?/2, which is the same election Oban.Stager gates on. Correct across a cluster, and needs an Oban instance running.

Choosing one

workflow do
  scheduler {AshWorkflow.Scheduler.Precise, leader: AshWorkflow.Scheduler.Leader.Oban}
end

Or with options for the implementation:

scheduler {AshWorkflow.Scheduler.Precise, leader: {AshWorkflow.Scheduler.Leader.Oban, name: MyApp.Oban}}

Single is the default because a workflow running on one node should not have to configure an election to get a timer. Precise logs a warning at boot when it is running on Single and more than one node is connected, since that is the configuration where deadlines fire twice.