What happened on Sunday

[IronQuarry48]

Thought this was worth its own thread.

What do you reckon?

[Violet Caitlin]

Sorry a mod installed broke I wasn't at home to be able to fix it

[Quanta]

Its working now

[Marcus]

What mod was it?

[Cole75]

A modified ProgressiveWeb app mods

[Hollow Ronan]

Something definitely felt off compared to previous Sundays. The timing of everything lining up with that update push makes me think it was less of a random outage and more of a cascading failure. Once one service stalled, everything depending on it followed.

The mention of a modified ProgressiveWeb app is interesting though. If they tweaked caching or service workers without a proper fallback, that could explain why some people were stuck even after things were "fixed". Seen that before where the client just refuses to let go of bad state

[Anchor34]

Not convinced it was just the PWA angle. That feels like a surface symptom rather than the cause. The scale people were reporting suggests something deeper, like backend auth or API rate limiting going sideways.

That said, PWAs can absolutely make recovery worse. If your app shell is cached incorrectly, users basically carry the bug around with them. Kind of ironic for something meant to improve reliability

[HollowFraction]

Funny thing is how split the reports were. Some people had zero issues while others couldn't load anything for hours. That usually points to regional infrastructure or CDN problems rather than a single bad deploy.

If it was a staggered rollout, maybe only certain nodes picked up the bad config. Would explain why clearing cache helped some but not others. Classic "works on my machine" at scale

[Lily98]

The whole situation reminded me how fragile these stacks can be. One small change in a dependency chain and suddenly half the platform is wobbling. Everyone builds on layers now, so when something breaks it's rarely isolated.

Also worth noting how quickly speculation fills the gap when there's no clear communication. A simple status update early on would've stopped half the guessing in this thread

[Forge40]

Bit of a tangent but this is why I still prefer having a proper fallback path instead of going all-in on fancy app behavior. When everything works, PWAs feel slick. When they don't, they can lock users into a broken state.

A basic server-rendered page might not be glamorous, but it usually still loads. Sometimes boring tech wins in moments like that

[Dank15]

The modified app theory makes sense if they changed how offline mode behaves. If the app thought it was offline when it wasn't, or vice versa, that could cause some really weird loops.

Still, the speed at which it spread suggests something centralized. One user glitch is a bug, thousands at once is infrastructure. Big difference there.

[ReplyGuy26]

Part of me thinks this was a monitoring failure as much as anything else. The issue seemed to go on longer than it should have before any visible response. Either alerts didn't fire or they underestimated the impact.

That's the kind of thing companies quietly fix after the fact, but it's usually where the real lesson is.

[CodyRhodes29]

Anyone else notice how mobile users seemed to have it worse? That lines up with the PWA angle since desktop browsers tend to be a bit more forgiving with cache invalidation.

Mobile apps wrapped as PWAs can get stuck in weird states where even reinstalling doesn't fully reset things. Bit of a nightmare to debug remotely

[Dialer75]

Could just be me, but the recovery felt uneven. Some features came back quickly while others lagged behind, which usually means different services were being restored independently.

That kind of staggered recovery can make it look like things are still broken even when the core issue is resolved. Not great for user confidence though.

[Gerrard98]

There's also the human factor. Weekend deployments are always risky, even if they're common. Smaller teams on call, slower response times, and fewer people available to triage properly.

Wouldn't be surprised if this was a routine change that just happened to hit at the worst possible time

[Ivory Molly]

People jumping straight to blaming one specific component feels a bit premature. Systems like this fail in messy ways, and root causes are often a chain rather than a single point.

Give it a few days and we'll probably hear a more nuanced explanation. Or we won't, and this thread will become the de facto postmortem