docs: memory leakage investigation report for SSR pipeline (#3296)

## Summary

- Adds a comprehensive investigation report for the reported memory leak
in react_on_rails_pro SSR workers (linear RSS growth under sustained
traffic)
- After building an in-repo reproducer and running 30,000+ requests over
68 minutes, **no reproducible heap leak was found** (1.63 KB/req
post-warmup, well below the 15 KB/req "refuted" threshold)
- The reported RSS growth is consistent with **memory fragmentation** —
a well-documented phenomenon in Ruby processes under glibc malloc with
large, variable-size allocations — not a simple heap-level object leak
- This type of growth cannot be reproduced in synthetic benchmarks
because it depends on specific allocator behavior (glibc malloc arena
expansion), traffic pattern diversity, payload size variation, and
runtime conditions that are rare in controlled environments
- Includes recommendations: jemalloc adoption, `MALLOC_ARENA_MAX=2`,
`--max-old-space-size` for Node workers, and worker rolling restarts as
the primary safety net

## Report Contents

1. **Investigation methodology** — reproducer infrastructure,
measurement approach, verification rubric
2. **SSR pipeline architecture** — memory allocation path from Rails
through HTTPX to Node Renderer
3. **Experiment results** — small payload (refuted), large payload
(inconclusive), 30K long run (refuted at 1.63 KB/req)
4. **Root cause analysis** — glibc malloc arena fragmentation, HTTP/2
amplification, CoW invalidation
5. **Why the leak cannot be reproduced** — deterministic data, uniform
concurrency, GC noise floor
6. **Application-level patterns** — common Node-side leak patterns in
user code (unbounded caches, saga middleware reuse, lodash memoize)
7. **Recommendations** — jemalloc, MALLOC_ARENA_MAX, heap snapshots,
monitoring guidance

Closes #3286

## Test plan

- [x] Report reviewed for accuracy against experiment data in
`MEMORY_LEAK_EXPERIMENTS.md`
- [x] No client-specific information included
- [x] All referenced files and code paths verified against current
codebase
- [ ] Stakeholder review of findings and recommendations

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Documentation**
* Added comprehensive investigation report documenting measurement
methodology, experimental findings across multiple configurations,
technical analysis of memory behavior patterns, and actionable
operational recommendations for optimization.

<!-- review_stack_entry_start -->

[![Review Change
Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/shakacode/react_on_rails/pull/3296)

<!-- review_stack_entry_end -->

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: AbanoubGhadban