Page loading fails after cache refresh

[LeMurphant]

It works once you refresh the page, but never if you don't.

[purpleidea]

I heard it was vibe coded, hence the bugs ;)

[LeMurphant]

@purpleidea not quite, Vibe Coding was not a thing when the UI was written a year ago. Now, look at any of my PRs, and there you will see Vibe Coding! That's what code review is for ;) (sorry @Aprillion)

[Aprillion]

I remember a long long time ago, cache used to work (relatively) fine... then people mobbed @mruwnik into implementing glossary and it's been a game of whack-a-mole ever since 😓 (that was a joke, the feature is useful.. but still, it's more caused by good-old feature creep and our collective tech debt than by vibe coding tech debt (for now))

Anyway, if you won't find what's the exact issue this time, feel free to deploy a testing CF instance with tons of debug code (or make a draft PR and I can deploy&debug if needed) ... I have suspicion that local miniflare might possibly behave differently around caching issues.

[LeMurphant]

Possibly related to #887

[Aprillion]

Notes:

• it can be sometimes reproduced on https://stampy-ui2.aprillion.workers.dev/questions/NM3T/ when reloading the page multiple times
• when GET /questions/allQuestionsOnSite returns HTTP 503 Worker exceeded resource limits but the main request as well as GET /categories/all return 200 .. the CPU limit for my free tier test worker is 10ms
• the main site was supposed to use paid tier with 30s CPU limit (and unlimited wall time), but when I checked Cloudflare dashboard, it shows "Free" in the Plan column 😱
  
• I don't have access to the owner account and I don't see Workers config when logged in member account
  => @plexish can you please check we are still using the paid tier for Cloudflare?

[Aprillion]

Updating dependencies doesn't seem to help - various medusa heads of this issue (HTTP 503, ...) are happening on https://stampy-ui2.aprillion.workers.dev/questions/NM37/AI-is-advancing-fast when using limited worker compute and when deployed from branch master...aprillion with updated libraries...

Next step might be to investigate performance of server-side parsing of responses from Coda, it everything was under 10ms maybe it would fix caching issues too (and if not, making it faster doesn't feel like it would have downsides).