in which bsky admits the outage 10 days ago *was* their fucking code after all.

[davidgerard@circumstances.run]

in which bsky admits the outage 10 days ago *was* their fucking code after all. they tried to blame AWS at the time and loudly insisted it wasn't their vibe-rotted code

https://pckt.blog/b/jcalabro/april-2026-outage-post-mortem-219ebg2
archive: https://archive.is/AKfKP

[theorangetheme@en.osm.town]

@davidgerard >Second, if you find this work interesting, we're hiring!

Lmao sure, let me just grab a steerage ticket for the Titanic at 2:19 AM on April 15th, 1912.

[klikini@pnw.zone]

@davidgerard
> o11y
It's time to stop

[klikini@pnw.zone]

@davidgerard Also, hilariously, I can't open any Bluesky links from that article because a request for app.bsky.ageassurance.getConfig is hanging!

[lemizy@defcon.social]

@davidgerard bluesky is vibe coded????

[laukidh@infosec.exchange]

@klikini @davidgerard they’re claiming this is a ddos, not more vibe code ‍️

[klikini@pnw.zone]

@Laukidh @davidgerard I wonder why anyone would DDoS that /s

[h3mmy@tech.lgbt]

@davidgerard They put synchronous loggers in high-volume transactions for memcache errors.

For context: That's like a textbook scenario for use non-blocking loggers.

- It's not a financial transaction that requires an airtight audit trail.
- High Throughput expected
- Production environment
- Prone to surges in traffic

The general pattern would be for the async logger to push it to a buffer and let a sidecar (like filebeat) scoop them up for the monitoring pipeline