Optimization loop:

1. Pick exactly one bottleneck to improve.
2. Make a single code change that targets that bottleneck.
3. Run the same performance test suite on every page.
4. Test conditions must stay identical every time:

   * same device
   * same browser
   * same network state
   * same cache state
   * same test method
   * same page order
5. Record the load time of every page.
6. Compare the new results against the previous run.
7. Keep the change only if:

   * every page is the same or faster, and
   * no page breaks functionally or visually.
8. If any page gets slower or regresses, revert the change.
9. Repeat from step 1.
10. Stop only when every page loads in under 50 ms, or when no safe improvement is left.

Decision rule:

* Improve
* Measure
* Compare
* Keep or revert
* Repeat

Output for each cycle:

* Change made
* Pages tested
* Previous timings
* New timings
* Best and worst page
* Decision
* Next bottleneck