docs: converge README Test servers section with main's restructure

[cliffhall]

Closes #1874

Step 2 of the ordered plan in #1821. Docs-only.

What was transplanted

main restructured the whole "Test servers" section in #1836; v2/main never got it. This brings that structure over — and only that section:

• New ### Serving the modern protocol era heading, with the transport.modern / ServerConfig options as a bullet list instead of a wall-of-text paragraph.
• New ### Showcase configs heading with a config → demonstrates → issue table indexing all nine showcase servers.
• Eight #### per-config subsections (MRTR, Network tab, x-mcp-header, page-by-page fetching, advertised extensions, logging, resource subscriptions, tasks), replacing the eight single long paragraphs v2/main still carried. Sub-tables for the MRTR presets and the trigger_* error taxonomy; numbered steps for the advertised-extensions walkthrough; legacy/modern bullet pairs for the era-fork configs.

No content was dropped — every test-servers/configs/*.json and behavior described on v2/main is still described, just under headings.

The #1846 correction is preserved

main's restructured copy still carries the stale claim. Kept main's blockquote-callout shape, corrected wording:

Before (on main):

Mcp-Param-* mirroring is skipped by the SDK in the browser (detectProbeEnvironment() !== "browser"). So calling get_weather from the web client omits Mcp-Param-City and the strict server answers -32020. The same tool is callable from the Node CLI/TUI, where mirroring is active.

After (this PR):

Mcp-Param-* mirroring is built by the Inspector, not the SDK. The SDK only mirrors inside client.callTool(), and skips it in the browser (detectProbeEnvironment() !== "browser"). The Inspector routes tools/call through client.request() to drive MRTR manually, so it builds the mirrored headers itself (#1846) — on every client, web included, since the web client's upstream request is issued by the Node backend rather than the browser. So get_weather is callable from web, CLI, and TUI alike, in both the plain and "Run as task" forms.

Not a whole-file copy

git checkout origin/main -- README.md was not used — it would have reverted v2/main-only content. The diff against v2/main is confined to lines 118–132, i.e. entirely inside the Test-servers section:

@@ -118 +118 @@   …   @@ -132 +143,97 @@

Nothing in #1866's regions (the "Repo status" callout near the top, the Project-layout tree, and the "Cutting a release" section) is touched, so this stays mergeable alongside it.

Residual diff vs main

git diff origin/main -- README.md now shows exactly three things, all expected:

1. The tools/call omits SEP-2243 Mcp-Param-* headers (x-mcp-header mirroring), rejected by strict modern servers #1846 correction above — the intended, deliberate divergence. This is the single 3-line hunk the first milestone merge will conflict on, where you take v2/main's side.
2. The quality-gate script table, prettier-reformatted on main — untouched here, merges cleanly (v2/main hasn't edited it since the merge base).
3. The trusted-publishing paragraph in "Cutting a release" — inside docs: post-swap branch model, PR policy, milestone rule, and Copilot review instructions #1866's region, deliberately left alone.

Payoff

The first v2/main → main milestone merge (#1821 step 5) drops from a whole-section README conflict to one 3-line hunk.

🤖 Generated with Claude Code

https://claude.ai/code/session_01YAt8rqxysNbhYWLhoRm3fU

[Copilot]

Pull request overview

Updates the repository’s README.md Test servers documentation on v2/main to match the restructured layout introduced on main, while preserving the corrected explanation of Mcp-Param-* header mirroring behavior (#1846 context).

Changes:

• Replaces a long “modern protocol era” paragraph with a structured “Serving the modern protocol era” subsection and bullets.
• Adds a “Showcase configs” index table and breaks the previous wall-of-text descriptions into focused per-config subsections.
• Keeps (and re-words) the Mcp-Param-* mirroring callout to reflect Inspector behavior rather than the SDK.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

Suppressed comments (1)

README.md:129

• The global guidance here says “unless noted, connect with Protocol Era = Modern”, but several listed showcase configs (e.g. pagination-http.json, advertised-extensions-http.json) do not set transport.modern and therefore can’t be reached when the client pins Modern (it will try to negotiate 2026-07-28 and fail). The intro should distinguish legacy-only configs from modern-serving ones so readers don’t hit a connection error following the instructions.

Each config below is a ready-made server for exercising one feature by hand. Load one with `--config`, and unless noted, connect with **Protocol Era = Modern**.

[cliffhall]

Round 2 of the Copilot loop came back clean — "reviewed 1 out of 1 changed files … generated no new comments". Round 1's only finding (the #1846 link pointing at /pull/1846, which 404s since #1846 is the issue and the fix shipped as PR #1847) is fixed in f7decc7.

Noting the one suppressed comment for the record, and why it is deliberately not addressed here:

The global guidance here says "unless noted, connect with Protocol Era = Modern", but several listed showcase configs (e.g. pagination-http.json, advertised-extensions-http.json) do not set transport.modern …

Fair point, and partly already covered — the advertised-extensions subsection does say "Legacy stateful leg only". But that sentence is verbatim transplanted text from main, not authored in this PR. Rewording it here would create a second divergence between the two copies of this section, which is exactly what #1874 exists to eliminate: the whole point is to get the first v2/main → main milestone merge down to the single intended 3-line hunk. Better addressed after the merge, when there is one copy of the section to fix.