@crowlKats ready for review

The WPT job is green — FileAPI/reading-data-section/filereader_events.any.{html,worker.html} now pass (both "empty blob" and "non-empty blob" subtests). The few red checks in the first run were unrelated CI flakes (Chromium/Playwright download failures in test specs on macos/windows-aarch64, and the flaky node_compat sequential/test-cluster-inspect-brk.js on windows-aarch64); I've re-run the failed jobs.

@crowlbot shouldnt we have a proper implementation of a task rather than settimeout? furthern investigate this spec language.

Thanks for the review — I looked into both points.

Spec language. The read operation says "queue a task" for each of loadstart / progress / load / loadend, so these are macrotasks — that's what the old queueMicrotask got wrong. The events are consumed one at a time, with the consumer re-arming for the next event in a microtask in between (the WPT EventWatcher does exactly this), so consecutive events have to land in separate event-loop turns for that re-arm to run before the next event fires.

Why setTimeout and not queueMicrotask/defer. The microtask-based deferrals don't give that separation. Two callbacks deferred in the same turn run back-to-back as microtasks — this is also true for op_defer/defer, since its .then reaction is a microtask — so a microtask scheduled by the first runs after the second:

setTimeout:     t1 → (microtask checkpoint) → t2     // ✓
queueMicrotask: t1 → t2   (no checkpoint between)    // ✗

Among the primitives we expose, setTimeout(fn, 0) is the one that queues each event as a real separate task. If you'd prefer a dedicated lightweight "queue a task" op in deno_core I'm happy to look into adding one, but I wanted to flag that the existing defer won't work here.

load + loadend. Good catch that the spec fires both within a single task. In the spec/browser model that task isn't author script, so a microtask checkpoint happens between the load and loadend listener invocations (the JS stack empties after each), which is what lets EventWatcher re-arm between them. We fire from a JS callback, so two synchronous dispatchEvents in one callback have no checkpoint between them. Rather than splitting loadend into its own task, I can keep them in one task (matching the spec wording) by awaiting a microtask between the two dispatches — I verified that also passes the WPT test. Happy to switch to whichever you prefer.