One task. Two dependencies.

A handled click.
A completed task?

Ask both versions for the same two shoes.
Watch what happens when the optional offer never arrives.

Browser simulation · no INP or speed score

1 Set the optional service

The simulated offer stays pending. Can the product task still finish?

2 Choose consent

3 Run the product task

Choose consent, then filter Shoes.

Choose for both versions. No preference is saved.

Preparing the browser simulation…

Waits for the offer

The product filter depends on an optional offer.

Click handler
Not yet
Correct result
Not requested
Optional offer
Not requested

6products in the catalog

  • Canvas trainer

    Shoes · €65

  • Day tote

    Bags · €40

  • Walking shoe

    Shoes · €90

  • Weekender

    Bags · €85

  • Sun cap

    Hats · €25

  • Wool beanie

    Hats · €30

Filters independently

The offer loads separately from the product task.

Click handler
Not yet
Correct result
Not requested
Optional offer
Not requested

6products in the catalog

  • Canvas trainer

    Shoes · €65

  • Day tote

    Bags · €40

  • Walking shoe

    Shoes · €90

  • Weekender

    Bags · €85

  • Sun cap

    Hats · €25

  • Wool beanie

    Hats · €30

The distinction

Feedback is only the beginning.

Start with a service that never answers. The same optional offer sits behind both versions.

Why measure both? In the frozen campaign, all 80 unfinished baseline tasks still recorded a 16 ms Event Timing duration. Those are archived results, not readings from this view. See the measured results.

The expected result is Canvas trainer, then Walking shoe, with no extra products.

What this comparison shows — and what it measures

Click handler means the handler ran and updated its feedback. It is not a measured Event Timing duration or proof of a painted frame. One button invokes the waiting handler first, then the independent handler, on the same main thread. Both use one shared offer promise; this is not a benchmark between independent pages.

Correct result comes from a separate DOM check: exact product IDs, names and order, with hidden/style/layout checks. Its own reads add work. A match must occur within 1,500 ms and still be present at the observation checkpoint. The checkpoint can run late in a background tab. The products and offer can still change afterward; the recorded checkpoint remains fixed. A changed or obstructed visual presentation is not fully covered by this DOM assertion.

Optional offer is simulated in this browser: a 600 ms timer, a rejected promise, or a promise held pending. Acceptance starts one shared simulation; rejection starts none. No offer request leaves your browser. A timer can fire later than its requested delay. If you wait before filtering, the offer may already be ready.

This teaching companion uses fresh, in-memory choices and separate presentation code. It does not reproduce the campaign’s network failures, input spacing, saved-choice states, CPU slowdown or measuring instrument. Its observations are never added to the frozen 480 trials. The original local laboratory uses two real HTTP origins, saved preferences and the unchanged measuring instrument. Keep its measured results separate from this browser simulation.