# Validation limits

This handbook targets MarkStackV2 revision 27a9763. Full pinned revision:

27a9763a7f1266a7495b825d818124862ae58255

Its package version is 1.1.4. The results below are scoped evidence, not universal performance, security, accessibility or production-readiness certification.

# Evidence levels

These levels do not substitute for one another. A layout screenshot cannot prove server authorization. A unit test cannot prove a production proxy forwarded the intended host/address. A local 404 cannot prove the external host does not return a homepage for missing pages.

# Pinned engine tests

An isolated copy of the upstream configuration passed all 219 executed tests on Node.js 24.21.0. The suite contains both structural/source assertions and executable behavior tests. It covers authentication, section access, stores, frontmatter/routes, search, local assets, editor state and publication fixtures at the assertions’ actual scope.

Test count is not feature count or exhaustive branch coverage. Expected failure fixtures can print errors while their assertions pass. Retain the actual command, runtime, configuration and final pass/fail summary rather than judging only log color or isolated error lines. Run your own checkout’s suite before a release; a retained result does not certify later code changes.

# Local handbook review

The handbook source tree has 80 Markdown pages. Its page/heading search-entry count changes with content revisions and is not the number of pages. Local checks verify generated internal link/fragment targets and search targets, while browser review checks selected pages and tasks at 320/390/768/1920 widths and in light/dark themes.

Phone-width emulation uses local Chrome, not physical iPhone Safari. The review does not claim all browsers or assistive technologies were tested. Internal-target checks do not validate every external URL, remote image, private audience or provider policy. Complete your own release and accessibility review for the deployed artifact.

# Historical timing sample

An earlier disposable draft had 79 pages and 357 page/heading entries. In that local environment, one build took about 0.77 seconds, Node-side index preparation about 19ms, and 1,000 repeated administrator queries about 2.15 seconds total. Those numbers describe the earlier fixture, not the current expanded handbook, browser first-load latency or organizational capacity.

Do not size a deployment from that sample. Measure actual source/output/index bytes, Node/browser/device versions, complete build time, first index fetch/preparation, repeated query latency and memory. Include a slow phone/network and actual diagrams. Static-reader distribution and editor write latency are different workloads; measure them separately.

# Unicode route fixtures

These were disposable builder fixtures for the pinned code. Route slugs and heading IDs use different rules. Empty/non-Latin titles can lose characters or collide; this is not a native multilingual routing or RTL feature. Use reviewed stable Latin route titles when needed, preserve non-English prose, inspect actual IDs/direction and test collisions. Do not silently transliterate the author’s intent or invent an accessibility guarantee.

# Recovery and service evidence

A simple starter source/config backup was restored into a fresh folder, compared with its original and built. That demonstrates a small source-recovery fixture, not recovery of production accounts, sessions, private runtime, service ownership or host cutover.

A systemd example passed local unit syntax verification. Syntax checking does not start the real service or prove permissions, bind address, reverse proxy, TLS, DNS or logging. Templates for systemd/nginx/Caddy/Apache and platform instructions need rehearsal on the actual installation.

# What is not certified

Record tested state and open gaps before release. Source-traceable documentation can explain a contract; it cannot create a feature, authority or operational guarantee the product does not implement.