chore(container-runner): tighten comments

[abcxff]

No description provided.

[abcxff]

Stack for rivet-dev/actors

• #5620 fix(container-runner): drain actors before announcing envoy …
• #5594 feat(container-runner): drain engine sleeps unless the idle …
• #5595 feat(container-runner): jitter idle timeout to avoid teardow …
• #5585 feat(container-runner): self-destroy on repeated actor start
• #5586 feat(container-runner): sleep on startup idle timeout
• #5589 chore(container-runner): tighten comments 👈
• #5587 feat(container-runner): exit process when the last child sto …
• #5588 feat(container-runner): drain child before SIGTERM on engine …
• #5584 feat(container-runner): drain children and engine concurrent …
• #5565 fix(envoy-client): ack terminating stop commands so pegboard …
• main

Get stack: forklift get 5589
Push local edits: forklift submit
Merge when ready: forklift merge 5589

_{change zrxxyxls}

[claude]

Review: chore(container-runner): tighten comments

Confirmed this is a comments/docs-only change — every hunk touches //////!/// doc comments, README prose, or comments in scripts/Dockerfiles. No behavioral code changed.

Finding: stale doc left inconsistent with the fix in this same PR (README.md:11-12)

The PR tightens container-runner/src/main.rs's module doc from "the instance stays warm after its last actor stops and never self-exits; the engine reaps it by draining the /start connection..." to "once the last child stops the process exits so the platform reaps the instance" — which actually matches the real behavior (stop_child in actor.rs calls request_exit when the child registry is empty, which cancels EXIT and lets main return/exit the process).

However, container-runner/README.md:11-12 still has the old, now-contradicted claim:

"The instance stays warm after its last actor stops; the engine reaps it by draining the /start connection after the request lifespan."

Since this PR is specifically about tightening/correcting these comments and already fixed the same inaccuracy in main.rs, it would be good to fix the README in the same pass so the two do not disagree.

Other observations (non-blocking)

• Em dash cleanup is consistent and complete — grepped all added lines and found no remaining em dashes, matching the CLAUDE.md rule against em dashes in prose/comments.
• The monitor.rs module doc trims out the old "memory/CPU sources in preference order" lists, but that detail is not lost — it now lives on the MemSource/CpuSource enum variant doc comments, a better place for it.
• The "Cloud Run" to "serverless"/"Rivet Compute" rewording is applied consistently across README, Dockerfile, docker-compose, and scripts. Good for not baking a specific vendor's product name into the code comments.
• One pre-existing em dash at README.md:187 ("— so thousands can run cheaply") is not touched by this PR and is out of scope, but since the PR is already doing comment cleanup in this file, it could be swept up too.

No functional, security, or test-coverage concerns — there is no logic change here, so no new tests are needed.

[claude]

Review

This PR is purely a comment/prose tightening pass over container-runner — I confirmed every hunk touches only ////// doc comments, the README, and script/Dockerfile comments; no executable code paths changed. Given that scope, risk is low, but a few things worth calling out:

Good

• Complies with the CLAUDE.md comment style rules: no em dashes remain in any added comment/prose (verified via grep across the whole diff), and the dash-parenthetical fragments from the originals (e.g. (if configured) - regardless of...-style phrasing) are gone in favor of complete sentences.
• The main.rs module doc change (removing "the instance ... never self-exits; the engine reaps it by draining the /start connection") actually fixes a stale doc, not just trims it — the branch stacks on feat-container-runner-exit-process-when-the-last-child-stops, and the new text ("once the last child stops the process exits so the platform reaps the instance") now matches actor.rs::stop_child, which calls request_exit when the child registry is empty. The old wording would have been actively misleading post-rebase.
• Similarly, dropping "failed start" from the EXIT static's doc comment ("actor stopped, failed start, or signal" → "last child stopped or a signal") is correct: on_start's error path explicitly does not tear down the instance ("A failed start is this actor's alone; it does not take down others").
• The monitor.rs module doc drops the detailed "memory/CPU sources in preference order" lists, but that detail is preserved per-variant on MemSource/CpuSource (src/monitor.rs:47-82), and the new comment explicitly points there (see [\MemSource`]/[`CpuSource`]`) — no net information loss.
• The Cloud Run → serverless/serverless platform terminology swap (README, Dockerfile, e2e scripts, main.rs) is applied consistently everywhere it appears, not just in the files that happened to get comment edits.

Minor / worth a second look

• A few trims do lose some genuinely load-bearing nuance rather than just wordiness. For example, in main.rs::async_main's exit-handling comment, the original explained why the signal path bounds the drain concurrently with notifying the engine ("Bounding the drain means an unreachable engine cannot eat the budget the children need") — that rationale is gone in the shortened version. Similarly drain_then_stop_child's doc drops the explicit tie-back to "the reclaim's SIGTERM→SIGKILL budget is honored." Neither is wrong, just thinner; worth confirming the team is fine losing that context in a low-traffic file like this one.
• Since this is a comment-only PR, there isn't (and doesn't need to be) new test coverage — nothing to flag there.

No functional, security, or performance concerns since no code semantics changed.