Bug tracking for SvelteKit — session replay, console + HAR (2026)

Svelte 5 replaced the magic of "assign to a variable, the UI updates" with **runes** — `$state`, `$derived`, `$effect`. The new system is more explicit and faster, but it changes how every Svelte bug manifests. SvelteKit 2 layered on top brings server load functions, form actions, and edge-deployed adapters — each with its own failure modes.

A SvelteKit bug report needs the route ID (because the same URL can hit different `+page.server.ts` files depending on layout), the form action result, and the rune state at the failure moment. BugMojo captures the DOM + console + network HAR end-to-end so the engineer can see exactly where the load function returned vs where the page rendered.

This guide covers Svelte 5 runes gotchas, SvelteKit 2 load-function bugs, and how to wire BugMojo into a Svelte team's workflow.

Framework-specific failure modes engineers actually ship through. Each is hard to spot in a screenshot and obvious in a session replay.

Real SvelteKit bug patterns — the symptom you will see in a report and the fix that actually works.

$state on an object does not deep-track new properties

Symptom: Adding a new key to a `$state` object does not trigger updates.

Fix: `$state` proxies via `new Proxy`. Properties that exist at declaration are tracked; properties added later need to be declared up front (set to undefined initially) OR use `$state.raw` and replace the whole object.

goto() before mount throws

Symptom: Navigation from `$effect` or `onMount` fails silently.

Fix: SvelteKit's `goto()` requires the navigation system to be initialized. Wrap in `tick()` or use `redirect()` from a load function instead.

Streamed promises in load() never resolve on the client

Symptom: A skeleton loader spins forever for a server-streamed promise.

Fix: Streamed promises require the response to remain open. A reverse proxy or compression middleware can buffer the response and break streaming. Test against a direct origin URL, then add infrastructure back layer by layer.

The honest comparison — where BugMojo wins, and where another tool might serve you better.