Operations | Monitoring | ITSM | DevOps | Cloud

You push a release that touches the checkout flow. Now you’re glued to dashboards and checking Slack, hoping you didn’t introduce a regression that breaks the payment path. You can’t tell if you’ve just shipped a blocker that’s stalling every cart—or some edge case quietly making users bail.

This is a guest post from Armin Ulrich, a fullstack developer, and founder of placid.app. He also created the MadeWith* network where he shares his projects and allows other developers to share theirs. There are many things I would rather do at 9pm than tracking down a mission-critical bug, but sometimes you don’t have a choice. Let me tell you the story about a slow query that led to a cascading failure–and how it could have been worse.

App hangs are the worst kind of bug: they don’t crash, they don’t log, and unless you're actively profiling, good luck catching them in the debugger. Maybe the main thread is blocked because it’s decoding a massive image with UIImage(data:). Maybe a background task is holding a lock or waiting on a DispatchGroup that never finishes. Maybe an async flow is stuck waiting on a continuation that never resumes.

You’re looking at an error in Sentry—a failed payment in your Flask backend or an unexpected null in your Node API. You’ve got the stack trace. The request details. Even the full trace. What you don’t have: the logs your app emitted right before everything went sideways. With Sentry Logs (now in open beta), you can send application logs straight to Sentry and see them automatically connected to the errors and traces you already use.

More code is being shipped today than ever before, accelerated by AI powered code gen tools. We’re in a golden age for builders. But here’s the thing: software still breaks in production. From a recent study by Microsoft, AI models struggle to debug software. It’s because most of these code gen tools lack the one thing every good developer relies on: context. To debug anything, you need context. Having AI tools doesn't change that.

So you're building an MCP server for your project or service, to allow AI chatbots and agents to interact with it? Great! You've decided to build it using Cloudflare Workers, have written the code, shipped it, and the first users are getting onboard: you're officially running it in production. That's when problems start. I'm not here to dissuade you from shooting your shot, but let's make sure you've got your bases covered in production when something inevitably goes wrong.