Config-only — purely additive entries to tests/node_compat/config.jsonc.
Verified each new line is alphabetically placed within its section (with the
ASCII collation the file uses, e.g. keep-alive-… before keepalive-…
because - < e, and setLocalWindowSize before settimeout-… because
L < t). All 9 entries pass that check.

Two of the additions (test-child-process-exec-any-shells-windows.js and
test-fs-realpath-on-substed-drive.js) are Windows-only and self-skip via
common.skip on non-Windows runners, so no platform gate is needed in the
config — the runner counts the skip as a pass. Worth keeping an eye on the
Windows runner specifically, since those two are the only ones exercising
real code paths there.

The other seven look low-risk to enable. The http2 batch
(max-concurrent-streams, server-settimeout-no-callback, socket-close)
is the area I'd watch most closely for after-the-fact flakiness given the
recent pattern in this suite — happy to mark/ignore if anything surfaces.
No production-code changes mean review surface here is small. LGTM modulo
post-merge CI watch.

Body lists 9 tests; diff has 4 net additions (the other 5 — test-child-process-exec-any-shells-windows, test-fs-realpath-on-substed-drive, test-http-keepalive-override, test-http-pipeline-assertionerror-finish, test-https-byteswritten — already on main via intermediate PRs, same merge-churn pattern as the recent batches).

Net additions checked alphabetically:

• test-http2-max-concurrent-streams.js
• test-http2-server-settimeout-no-callback.js
• test-http2-socket-close.js
• test-tls-startcom-wosign-whitelist.js

Holding off on APPROVE — CI has two red jobs but neither is actually caused by this PR's changes:

1. lint debug windows-x86_64 — fails with Error: .github/workflows/start_release.generated.yml is out of date. Run: .github/workflows/start_release.ts. That's a workflow-regen check that's stale because this branch hasn't picked up a recent main commit that touched start_release.ts. A merge from main and re-running the generator will fix it; nothing for this PR's actual diff to address.

2. test node_compat (2/3) release linux-x86_64 — exit code 8 during sysroot setup (download/extract from deno_sysroot_build), well before any test ran. Pure infrastructure flake; a CI re-run should clear it.

Per fibibot's new CI gate I can't autonomously APPROVE while there are red rows, even when the failures are infrastructural. Quick fix is to rebase on main + click "re-run" on the linux job. Once those land green I'll re-confirm.

131/134 passing checks, all relevant test node_compat debug and test unit_node debug are clean.