Differences in pendingMs between loader and suspending

[DogPawHat]

One of the things that I came across when implementing the Unified Imperitive Methods for Query, is when intergrating TanStack Query (or any other external data loading library for that matter) into router, when should you should await promises in the loader in the first place.

There's a pattern where you can rely on the route-level suspense and error boundaries to trigger the pendingComponent instead of awaiting in the loader; Dominik gave a very good example of this pattern here: https://tkdodo.eu/blog/tan-stack-router-and-query#to-await-or-not-to-await-in-the-loader

Heres the nitpick with this: If you suspend in a component and trigger pendingComponent, it will ignore pendingMs and show the pendingComponet immediately. This was something I've been aware of since I raised it back in the Query RFC at TanStack/query#9135 (reply in thread) but it obviously needs a proper discussion here in router.

Is always showing the pendingComponent/defaultPendingComponent when a component suspends and gets caught by the route boundary the intended behaviour, regardless of what pendingMs says? I'm not the biggest fan of it myself (I would agree with Domninik here and just set it to 0) but if someone was to try and use it in conjuction with the "never await, always suspend" pattern it's a little unintuitive.

I've set up a demo repo to show the difference between the two modes here: https://github.com/DogPawHat/pending-ms-tests-tanstack . Deployed it on netlify: https://exquisite-melomakarona-9128a5.netlify.app/

[schiller-manuel]

Is always showing the pendingComponent/defaultPendingComponent when a component suspends and gets caught by the route boundary the intended behaviour, regardless of what pendingMs says?

are you asking about SSR here?

[DogPawHat]

No, the issue occours without ssr and with Start's SPA mode on.