Event triggers versus schedules: separate timely processing from reconciliation
Compare change-triggered execution with periodic polling, checking duplicates, omissions and time windows.
Choose an approach
Consider events for timely changes with suitable sources and schedules for periodic work or missing events. Combine events with reconciliation. A trigger is not completion and schedules do not guarantee coverage. Define identifiers and windows and check final states.
Separate start conditions from pending work
Power Automate distinguishes event-triggered and scheduled flows; triggers begin work but do not complete it. Sources may repeat notifications and scans may find still-running records, requiring system-specific checks. Run counts are not new-task counts. Define created-versus-updated time windows and processed identifiers to avoid repeats or missed transient records.
Test duplicates and cross-window changes on one record set
Use fixed-ID records for both approaches. Repeat an event and check the expected single effect; update a record near a window boundary and verify reconciliation. Resume unfinished work after interruption rather than blindly restarting. Separate trigger time from completion. No real schedule or external system was changed and no delivery or punctuality rate is given.
Long-running automation needs reconciliation and recovery
Recheck event conditions after interface changes and schedule time zones and windows. Preserve failures and completion IDs for reconciliation. Count empty polls, duplicates and backlog recovery. Simple schedules can suit low volume; timely requirements may justify events. Choose around data changes and trace individual completion rather than scheduler uptime alone.
Checked sources
These sources support the concepts and methods described here. Examples and test scope are labeled separately; check the version of the tools you use.