docs(paper): refresh abstract to completed-DAG story for ASE 2026 Industry Showcase (TIN-2106)

[Jesssullivan]

What this PR does

Refreshes the abstract and stale prose in docs/paper.md (paper A, the ASE-target experience report) from the in-progress "feature-flag strangler fig" framing to the completed-DAG reality shipped in bridge 0.7.0, and adds an operator-facing submission packet. This is for operator review of scholarly content — see the DO-NOT-MERGE/SUBMIT note below.

Changes to docs/paper.md

• Title -> Typed Browser Automation as an Anti-Corruption Layer: A FOSS Offramp Pattern for SaaS Vendor Lock-in.
• Abstract replaced with the fact-checked final abstract (experience-report framing; completed deletion gate; derived-projection flow DAG; anti-renaming golden-trace conformance; per-flow R-typing transport-agnosticism; two auth-free Bazel-vendored adopters).
• Contributions list refreshed: adds the flow DAG as a derived FlowPlan projection (sha256 plan hash), the anti-renaming deletion gate (538 insertions / 1,305 deletions / 16 files), and compile-time transport-agnosticism via per-flow R typing.
• Intro / Discussion / Conclusion: reframe the bridge-internal strangler as completed (three divergent compositions, migration feature flag, kill switch, and shadow mode deleted; single ~100-line runFlow fold the sole path). Kept the application-level Acuity-vs-homegrown selector as a separate axis (lives in the MassageIthaca app), with an inline OPERATOR-VERIFY note so the bridge-level completion is not overclaimed as a vendor sunset.
• Deployment section: Modal-only -> "originally Modal, now a self-hosted Kubernetes (RKE2) cluster consuming the same Bazel-built container artifact"; Node 22 -> Node 24; Bazel //:pkg build authority.
• Preserved the real numbers exactly: 604 appointments, 62 weeks, 8 blocked endpoints, production since March 2026, 17-method interface.
• Inline <!-- OPERATOR-VERIFY: ... --> comments flag the 3 spots needing operator confirmation.

New file docs/paper/ase-2026-abstract.md

Operator submission packet: the final abstract verbatim, the ASE Industry Showcase CFP requirements summary, and the full claim -> status -> evidence fact-check table (26 rows, every empirical claim traced to code / git / Linear / the existing paper).

ASE 2026 Industry Showcase CFP (summary)

• Format: ACM acmart sigconf,review (NOT IEEE), PDF. Single-blind — author NOT anonymized.
• Length: short 5+1 (recommended, experience report) or long 10+2.
• New mandatory 2026: Data Availability Statement after Conclusions pointing to a public DOI-archived artifact [OPERATOR-VERIFY DOI].
• Editorial bar: impact + real-world evidence over novelty; marketing/PR excluded.
• DEADLINE FLAG: the confirmed ASE 2026 Industry Showcase dates are abstract 23 Apr 2026 / paper 30 Apr 2026 (AoE) — both already past; HotCRP shows submissions CLOSED. The assumed "2026-06-23" date does not match this track. The operator decides the target cycle (a still-open ASE 2026 track, the ASE 2027 cycle, or contacting the chairs). Requirements are stable year-over-year, so this refresh carries over with near-zero rework.

Fact-check outcome

All 26 abstract claims verified against the cloned repo @ main (HEAD 0c7ec4b, 0.7.0): deletion-gate stat is an exact match (538 ins / 1,305 del / 16 files via git show --stat 5ca57ed); runFlow is the single fold; FlowPlan carries a sha256 hash; the CalCom r-channel-proof.ts source fence is real; the migration flag / kill switch / shadow mode are deletion-only comments. No contradictions with the preserved numbers. Framing caveats (both adopters operator-affiliated; each adopter's live production behind a manual promote; bridge-level vs application-level strangler) are captured in the fact-check table and inline OPERATOR-VERIFY comments — the abstract stays accurate ("consume the stack end-to-end", not "in production").

Build

• The distinct companion paper docs/paper/acuity-middleware-paper.tex (the SETNX / O(1) single-flight work — a different paper, not a twin of paper A) was not overwritten; the two abstracts are different works. It still builds cleanly with tectonic (pnpm paper:build, exit 0, ~129 KiB PDF).
• docs/paper.md is Markdown (no build). mkdocs build --strict passes; pnpm docs:check passes.

⚠️ DO NOT MERGE / DO NOT SUBMIT until operator review

This is the operator's scholarly output. Submission is the operator's act — do not email, post to any portal, or submit anywhere. Merge only after the operator reviews the scholarly content and resolves the OPERATOR-VERIFY items (incl. the DOI for the Data Availability Statement and the closed-deadline target-cycle decision).