Back to the changelog

v0.52.3

July 18, 2026

A faster engine, same behavior: every state transition makes far fewer database round-trips and pre-transition checks run in parallel; the alert engine drops repeated queries and always prioritizes the longest-waiting work.

  • Improvement

    Faster state transitions

    When entering a new state, the engine loads the status once and shares it across data gates, events and automatic rules (it used to re-read it up to 5-6 times), loads flow variables only when the timeout expression actually uses them, and runs pre-transition checks and rule-cycle reads in parallel. Same results, same error messages: just faster.

  • Improvement

    A more efficient and fair alert engine

    The waiting-work scan resolves assignees with a single cumulative query instead of one per candidate, and alert evaluation reads task states in bulk. On top of that, scans always process the longest-waiting items first: even under heavy backlog no alert is left behind.