docs(research): VMAFX eBPF optimization target + implementation plan

[lusoris]

Summary

Research-0733 surveys the VMAFX Phase 4b workload hot paths for eBPF optimization
(ADR-0709, item 4b.6) and selects rclone FUSE page-cache bypass via eBPF kprobe
on fuse_file_read_iter as the highest-value target.

• Projected 15–40% job wall-time reduction on warm-cache nodes for 1080p60 clips
• 37× p50 FUSE read latency reduction (from ~150 µs to ~4 µs for cached reads)
• Four-phase implementation plan documented (4b.6.a baseline → 4b.6.d Helm chart)
• Research-only PR; no eBPF code written in this PR

Type

• docs — documentation only

Checklist

• Commits follow Conventional Commits (the commit-msg hook enforces this).
• make format && make lint is green locally.
• Unit tests pass: meson test -C build.
• If I touched any SIMD/GPU code path, I ran /cross-backend-diff and the worst ULP is ≤ 2.
• If I touched a feature extractor with SIMD/GPU twins, I either updated every twin or listed the gap under "Known follow-ups" below.
• If I added a new .c / .cpp / .cu / .h / .hpp, it has the appropriate license header (see CONTRIBUTING.md).
• If this is a breaking change, the commit message uses ! or BREAKING CHANGE: and the migration path is documented below.
• If this PR adds an ADR, the ADR row lives in docs/adr/_index_fragments/<NNNN-slug>.md and the slug is appended to docs/adr/_index_fragments/_order.txt — do not edit docs/adr/README.md directly (regenerated by scripts/docs/concat-adr-index.sh; see ADR-0221).

Bug-status hygiene (ADR-0165)

• docs/state.md updated in this PR with a row in the appropriate section — T-VMAFX-EBPF-RESEARCH-4B6-2026-05-28 added to Recently closed.

Netflix golden-data gate (ADR-0024)

• I did not modify any assertAlmostEqual(...) score in the Netflix golden Python tests.
• If I believe a golden value must change, I have explained why below AND pinged @lusoris for a CODEOWNERS exception.

Cross-backend numerical results

No code changed; no cross-backend diff required.

Performance (if perf or feat)

No code changed; projected improvement is documented in Research-0733 §4.5 and §8.

Deep-dive deliverables (ADR-0108)

• Research digest — docs/research/0733-vmafx-ebpf-optimization-target.md written.
• Decision matrix — captured in Research-0733 §3 (candidate scoring) and §6 (alternatives considered).
• AGENTS.md invariant note — no rebase-sensitive invariants: docs-only PR; no build system or C source changed.
• Reproducer / smoke-test command — see below.
• CHANGELOG fragment — changelog.d/changed/ebpf-research.md added.
• Rebase note — entry added to docs/rebase-notes.md: no rebase impact for this docs-only PR.

Reproducer

# Verify the research digest is well-formed and cross-references are intact:
grep -c "fuse_file_read_iter" docs/research/0733-vmafx-ebpf-optimization-target.md
# Expected: >= 5 (the kprobe symbol is referenced in §2.1, §4.1, §4.2, §5, state.md row)

Known follow-ups

• Phase 4b.6.b: feat(ebpf): fuse-latency eBPF probe standalone prototype
  — pkg/ebpf/fuse_latency.bpf.c + pkg/ebpf/fuse_latency.go + cmd/vmafx-ebpf-probe/main.go
• Phase 4b.6.c: feat(ebpf): integrate FUSE latency monitor into vmafx-node
• Phase 4b.6.d: feat(deploy): Helm chart eBPF privilege grant for FUSE latency monitor

🤖 Generated with Claude Code