v2.1.36

Pull Requests

• test(F-1000): the domcontentloaded ordering pin holds load on the fixture's gate instead of reading the runner's clock (#239)
• test(F-1001): the lost-port pin requests a port the OS never assigns, so an ephemeral bind cannot collide with it (#240)
• fix(F-1002): networkidle's fixed sleep leaves the landing read half of a small budget; the networkidle pin has its own deadline (#241)

Commits

• 8cc1969 release: 2.1.36
• 66c8ec5 test(F-1003): the spawn-phase bound pin no longer says three phases
• 0328d5a docs(F-1003): review - the spawn budget now has four phases; first_page is its own phase, not inside the post-launch one
• 0d8f952 fix(F-1003): review - wait for the delegated launch too, bound the page wait by the post-launch spawn budget
• 0e82508 fix(F-1003): a launched Chrome that has not opened its first page yet is waited for, not killed
• 9379dea review(F-1006): a recovered slow settle warns that navigate returned early (T-4709); expect_app rejects a page without app.js
• 9a01526 review(F-1006): the ready wait needs a complete document, tolerates a failed probe, names what it saw; pinned
• 00503fe test(F-1006): the harness waits for the fixture app to be wired before it interacts
• b767977 review(F-1005): scoped's annotate answers VALUE only, so STRING/FORWARDREF stay real
• 49ec22f fix(F-1005): the backend registers its tools on Python 3.14
• 3221894 review(F-1004): raise a delegated Chrome's existing children too; nothing escapes the raise; pin once, dead-pid and lifespan order behaviourally
• 6058171 fix(F-1004): a backend started by Task Scheduler no longer hands Low I/O and memory priority to every Chrome it launches