Heuristic

Start with what changed, not with certainty about it

The last change deserves an interview, not an automatic conviction.

When it fits

  • A previously stable system suddenly changes behavior.

When to avoid it

  • Failures can emerge without a local deployment when dependencies, data, capacity or external conditions change.

Why it matters

Build a short list of deployments, configuration changes, dependency changes, data-shape shifts and load changes near the onset of the symptom. Compare their timing with the first observable deviation. Use the list to prioritize tests, but require evidence that connects a change to the failure before calling it the cause.

An example

A job starts timing out after a transport import; compare the failure window and behavior before assuming the transport is responsible.

Check your result

At least one recent-change hypothesis becomes stronger or weaker because of a test, not because of chronology alone.

Keep this limit in mind

  • Failures can emerge without a local deployment when dependencies, data, capacity or external conditions change.

Connected ideas

Useful with
Prefer a test that kills hypotheses

Evidence and sources

Supports

Google SRE identifies recent deployments, configuration changes and environmental shifts as productive leads when troubleshooting sudden changes in behavior.

Temporal proximity is a lead, not proof of causation; an unchanged component can still fail because its inputs or dependencies changed.

Effective Troubleshooting · What touched it last

All sources (1)