Minimal Repro Bundle

How It Works

This recipe guides you through producing a minimal reproducible example (MRE) or,

when that fails, capturing the right diagnostics to make the next occurrence debuggable.

Triggers

• "Works on my machine" / "Cannot reproduce"
• Bug occurs only in CI, staging, or a customer environment
• Intermittent failures with unclear root cause

Steps

1. Capture the reproduction story:

• exact inputs,
• expected vs actual behavior,
• frequency and environment details.

1. Reduce to an MRE:

• smallest dataset,
• smallest code snippet,
• single command to run.

1. If not reproducible: instrument strategically:

• add structured logs around suspected boundaries,
• add correlation IDs and version info,
• capture config diffs (non-secret).

1. Produce a "debug bundle":

• env versions,
• config flags,
• logs/traces,
• steps to run.

1. Add a regression test once fixed.

Expected Outcome

• Either a concrete repro you (or others) can run, or a clear list of missing observability.
• Less back-and-forth with QA/customers and fewer stalled tickets.

Example Inputs

• "Customer sees a crash; I can't reproduce locally."
• "CI fails intermittently; local always passes."
• "Bug only happens with specific OS/browser/version combinations."

Tips

• When you can't reproduce, assume an instrumentation gap, not user error.