DEV Community

David Lyon
David Lyon

Posted on

AWS Confirmed the Crash. Then Denied It. Then Billed Us.

In May, I published a detailed technical breakdown of a fatal, platform-level crash in AWS Lambda — a Node.js function inside a VPC making outbound HTTPS calls that would silently crash after returning a 201.

No logs. No errors. No stack trace. AWS eventually reproduced it themselves.

This post isn't about the bug.

It's about what AWS did next.

What Happened After the Reproduction

Even after AWS reproduced the crash using their own test code, their response wasn’t engineering.

It was damage control:

  • They blamed our code
  • They revoked a previously granted $4,000 credit
  • They silently billed our founder’s personal card
  • They filtered technical escalations through sales
  • They offered the credit back — this time as a settlement tied to silence
  • And they asked our CEO, off-record, not to publish

We published anyway.

The Real Postmortem: Culture, Not Code

This isn’t a teardown of runtime behaviour. That’s already done.

This is about support failure, accountability avoidance, and AWS’s cultural instinct to deflect rather than own fault.

If you’re building serious workloads on Lambda: read this.

Because when support fails, logs go dark, and engineering won’t speak — you are on your own.

Read the full follow-up post here

Top comments (2)

Collapse
 
m_azarboon profile image
Mahdi Azarboon • Edited

The link to follow up post, doesnt work. Seems the whole website is down

Collapse
 
rosswilliams profile image
rosswilliams

I don't see your minimal reproduction of the issue. Can you post the reproduction and your expected output?