[robjtede]

Move only Lint / fmt (no matrix) to ubuntu-slim. This job checks justfile, TOML, and Rust formatting without compilation. All steps and permissions remain unchanged.

Retain CI's platform/toolchain test and build matrix, Clippy, rustdoc, and instrumented coverage on their existing runners because they compile or process Rust dependencies. Retain release automation, which can build/publish packages. Read MSRV uses a pinned upstream reusable workflow with its own runner selection.

GitHub documents slim availability for public/private repositories, x64, 1 CPU, 5 GB RAM, 14 GB storage, and a 15-minute hard job limit. It runs in an unprivileged container: Docker-in-Docker, mounts, and low-level kernel access are unsupported. Its minimal image includes shell, Node, Git and download utilities; Rust is installed explicitly by this job. Sources: https://docs.github.com/en/actions/reference/runners/github-hosted-runners and https://github.com/actions/runner-images/blob/main/images/ubuntu-slim/ubuntu-slim-Readme.md

Validation: just check-ci passed locally. Remote slim compatibility and timing remain pending. No release or coverage workflow was manually triggered.

Timing evidence: historical successful Lint / fmt, ubuntu-latest, n=3. Execution is completed_at minus started_at; queue time is excluded and has not been measured.

Run	Commit SHA	Seconds	Result
https://github.com/bytesize-rs/bytesize/actions/runs/41382128375	5bce216	21	success
https://github.com/bytesize-rs/bytesize/actions/runs/41382120024	56c3157	22	success
https://github.com/bytesize-rs/bytesize/actions/runs/41382114917	02812aa	18	success

Baseline median: 21 seconds. First sample used Ubuntu 26.04 and had a Rust cache hit; other cache states are not yet inspected. Historical comparisons are confounded by different commits, image versions, nightly toolchains, and cache state. Slim n=0, median and 100*(slim/base-1) unavailable. No compatibility failures observed yet; remote testing is pending.

Measurement update: first slim run passed, with all 7 required checks passing on db2a197. Lint / fmt (no matrix), ubuntu-slim, run https://github.com/bytesize-rs/bytesize/actions/runs/41636841355, job 125920067317: 50 seconds (12:34:39–12:35:29 UTC). Slim n=1 median 50s versus historical baseline n=3 median 21s: +138.1%. This is not an isolated runner-performance estimate: different commits/images/nightlies and cache conditions apply; slim downloaded crate dependencies during cargo metadata. Queue time excluded and unmeasured. No compatibility failures. A formatting-only rerun is requested for another sample; release/publishing jobs are not triggered.

Final available measurement update (supersedes pending statements above): attempt 2 formatting passed in 29 seconds (12:38:00–12:38:29 UTC), with a confirmed Rust cache hit and successful restore: https://github.com/bytesize-rs/bytesize/actions/runs/41636841355/job/125921777859 . Same head db2a197, ubuntu-slim, Lint/fmt, no matrix. Slim samples [50,29], n=2, median 39.5 seconds; historical baseline [21,22,18], n=3, median 21 seconds. Median change = 100*(39.5/21-1) = +88.1%. The first slim sample populated the cache, the second restored it. Mixed cache conditions and historical baseline commits prevent attribution to CPU performance alone. Queue time remains excluded/unmeasured. No additional benchmarks were started for the failure-status request.

Failure audit: all 10 current checks are successful. CI attempt 1 and Lint attempts 1 and 2 have no failed, timed-out, or cancelled jobs. No earlier resolved failures, capacity cancellations, or rerun API rejections occurred in this PR assessment. Clippy/docs shown on Lint attempt 2 are carried-forward original successes, not fresh execution samples. No compatibility failure was observed. No measurements remain running; third slim sample and same-code baseline were not collected.