Web monkey traffic collection with persistent browserContext

[baltpeter]

I've been meaning to run a web traffic collection with a persistent browserContext for a while, working under the assumption that we miss out on a fair bit of tracker behaviour and data transmissions if the profiles they can collect are limited to one minute on a single site.

I haven't implemented this yet due to the way Playwright handles HAR recording. So far, you could only create a HAR by starting and stopping a browserContext (https://playwright.dev/docs/next/api/class-browser#browser-new-context). This way, it's not possible to reliably determine the initiatorUrl.

[baltpeter]

I just started looking into this again. I considered using https://www.npmjs.com/package/playwright-har, which does appear to provide the desired API, however it has two downsides:

• It only works with Chromium-based browsers due to being based on CDP (which would be fine, First traffic collection on the web #3 used Chromium anyway).
• It has a custom HAR recording implementation (with limitations) that appears to not been maintained in a while (less good).

[baltpeter]

However, we seem to be in luck. Just a week ago, a PR was merged that adds new tracing.startHar and tracing.stopHar methods (https://playwright.dev/docs/next/api/class-tracing#tracing-start-har) that on first glance look like just what we need!

Looks like these new APIs will be released in Playwright 1.60, which isn't released yet. But there are alphas (https://www.npmjs.com/package/playwright/v/1.60.0-alpha-2026-04-20) we can test. Let's give this go.

[baltpeter]

Seems to be working well.

[baltpeter]

It's probably not fair to the sites to distinguish between no-interaction and monkey in this setup, though, since we may have previously visited the site and consented.

I'll call everything monkey but wait some amount of time before starting the monkey to still capture the pre-interaction behaviour.

[baltpeter]

I'll go with the top 1k as per CrUX again.

BigQuery query:

SELECT DISTINCT origin, experimental.popularity.rank FROM `chrome-ux-report.country_de.202603` WHERE experimental.popularity.rank <= 1000

Dump: bquxjob_677a17dc_19dab26723d.json, bquxjob_677a17dc_19dab26723d.csv