docs: add measured RSC results

Author: tannerlinsley

rsc-migration-report.md

@@ -1,6 +1,6 @@
 # RSC Migration Report
 
-Status: pre-ship bundle + code audit complete. Current production Lighthouse baseline captured. Post-deploy production compare still pending.
+Status: bundle + code audit complete. Production Lighthouse baseline and post-deploy compare captured.
 
 ## Scope
 
@@ -160,13 +160,11 @@ Notes:
 
 ## Post-Deploy Compare
 
-Fill this in after deploy to `main` and re-run Lighthouse on the same production URLs.
-
 | Page                            | Before Score | After Score | Before LCP | After LCP | Before TBT | After TBT | Before Bytes | After Bytes | Notes |
 | ------------------------------- | -----------: | ----------: | ---------: | --------: | ---------: | --------: | -----------: | ----------: | ----- |
-| `/query/latest`                 |           80 |             |       3.4s |           |      300ms |           |      765 KiB |             |       |
-| `/blog/react-server-components` |           52 |             |       3.7s |           |    1,200ms |           |    1,101 KiB |             |       |
-| `/router/latest/docs/overview`  |           78 |             |       3.6s |           |      280ms |           |      917 KiB |             |       |
+| `/query/latest`                 |           80 |          77 |       3.4s |      3.8s |      300ms |     250ms |      765 KiB |      784 KiB | Roughly flat. Slightly worse score/LCP on this run, slightly better TBT. |
+| `/blog/react-server-components` |           52 |          74 |       3.7s |      3.6s |    1,200ms |     260ms |    1,101 KiB |      785 KiB | Clear win. Huge TBT and transfer-size drop, major score improvement. |
+| `/router/latest/docs/overview`  |           78 |          81 |       3.6s |      3.6s |      280ms |     200ms |      917 KiB |      777 KiB | Modest win. Better score, lower TBT, lower transfer size. |
 
 ## Conclusions So Far
 
@@ -175,3 +173,25 @@ Fill this in after deploy to `main` and re-run Lighthouse on the same production
 - Clear removal of markdown parsing and syntax highlighting from the client bundle
 - Not a universal bundle win across all landing pages yet
 - Best pages to watch post-deploy: blog and docs
+
+## Post-Deploy Read
+
+- Blog got the clearest real-world benefit.
+  - Lighthouse score improved from `52` to `74`
+  - TBT dropped from `1,200ms` to `260ms`
+  - transfer size dropped from `1,101 KiB` to `785 KiB`
+- Docs improved, but less dramatically.
+  - Lighthouse score improved from `78` to `81`
+  - TBT dropped from `280ms` to `200ms`
+  - transfer size dropped from `917 KiB` to `777 KiB`
+- Query landing did not show a Lighthouse win on this run.
+  - Score went from `80` to `77`
+  - LCP worsened slightly from `3.4s` to `3.8s`
+  - TBT improved slightly from `300ms` to `250ms`
+
+Interpretation:
+
+- The bundle-side story predicted this fairly well.
+- The strongest wins landed on markdown-heavy pages where we removed the client markdown/highlighting pipeline.
+- The landing-page story remains mixed because those pages were not dominated by markdown and some of the server-rendered example-shell work traded complexity reduction for slightly more initial shell work.
+- If we want broader Lighthouse wins on library landings, the next work is probably not more RSC conversion. It is targeted landing-page performance work.

src/blog/react-server-components.md

@@ -449,6 +449,40 @@ You choose patterns **per-route, per-component, per-use-case**. The architecture
 
 ---
 
+## What We Saw On TanStack.com
+
+We did not want to make the case with vibes, so we migrated the content-heavy parts of tanstack.com and measured it.
+
+The result was exactly what we hoped for, and also more limited than the hype would suggest.
+
+The best pages got meaningfully smaller:
+
+- **Blog post pages** dropped about **153 KB gzipped** from the client JS graph.
+- **Docs pages** dropped about **153 KB gzipped**.
+- **Docs example pages** dropped about **40 KB gzipped**.
+
+And the real-world numbers moved with them:
+
+- **`/blog/react-server-components`** went from **52 -> 74** in Lighthouse.
+  - Total Blocking Time dropped from **1,200ms -> 260ms**.
+  - Transfer size dropped from **1,101 KiB -> 785 KiB**.
+- **`/router/latest/docs/overview`** went from **78 -> 81**.
+  - Total Blocking Time dropped from **280ms -> 200ms**.
+  - Transfer size dropped from **917 KiB -> 777 KiB**.
+
+That is the point. **Heavy client work stopped shipping to the client.** Markdown parsing went away. Syntax highlighting went away. The browser got less JavaScript and did less work. As a side effect, we also got to delete the old client markdown and highlighting path instead of carrying two versions of the same rendering logic around.
+
+But RSCs are not a universal coupon code for performance. Some landing pages were basically flat, and a few were slightly worse. Pages that are already dominated by interactive UI shell do not automatically get faster just because you used a server component somewhere in the tree.
+
+That is the tradeoff:
+
+- **RSCs are great when the page is content-heavy, dependency-heavy, or both.** Docs, blogs, markdown pipelines, syntax highlighting, slow-changing content, SEO-heavy pages. That is the sweet spot.
+- **RSCs are less obviously useful when the page is already mostly client state and interaction.** Dashboards, builders, long-lived app sessions, and some landing pages can be flat or mixed unless you are removing real client-side work.
+
+That is why we think they matter. Not because every route should become a server component. Because when you use them where they fit, the payoff is measurable and not subtle.
+
+---
+
 ## Current Status: Experimental
 
 RSC support is experimental in TanStack Start RC and will remain experimental into early v1.