In real-world distributed systems, network jitter, collector restarts, and ephemeral containers frequently cause datapoints to arrive late or go missing.
SignalFlow relies on two critical controls to handle irregular telemetry: max_delay (or maxDelay, configured in Chart/Detector options or API execution settings rather than inside data()) and extrapolation policies. The max_delay setting tells the streaming engine how long (for example, 30s or 5m, or auto) to wait for slow datapoints before running calculations for a given timestamp. Setting max_delay too short drops late-arriving metrics; setting it too long adds latency to charts and alerts.
Extrapolation defines how Splunk treats missing intervals inside an active MTS when aggregating across a fleet: Null (default/strict) leaves gaps uncounted, Last Value holds the most recent reported number, and Zero treats missing intervals as 0 (useful for sparse error counters that only emit when errors occur).
# max_delay / maxDelay is configured in Chart/Detector/API execution settings, not inside data()
errs = data('payment.errors', extrapolation='zero', maxExtrapolations=3)
errs.sum(by='region').publish(label='Payment Errors')