🧪 Make lifecycle and publication tests observe complete outcomes

[taras]

Why

Several passing tests do not observe the complete outcome their names promise: the REPL teardown case reads a counter copied before unwind, concurrent replacement accepts a shared text prefix, and release tests compare artifact and target names without retaining their association. This first slice of #914 makes those gaps observable.

What changes

Before: cleanup, independent subscriptions, exact publication bytes and original failure status can be missing while the selected assertions still pass. FE19's description also advertises an obsolete input alias that its programs never use.

After: tests observe the live post-unwind counter, gated producer cleanup, two producers from two subscriptions to one stream, complete submitted replacement bytes, exact release target/artifact pairs, the first failing file's original status, and exact weights replacing a preexisting file. FE19 preserves canonical text repeatability and refuses source/program before a live fragment-read sentinel fires.

What must stay true

• Journal stability remains separate from the REPL release count.
• Later authored work stays unreached, independent expansion-call evidence stays separate, and each subscription returns its own output.
• FE21 retains real Workspace routing; the current text-only product contract remains unchanged.
• Release membership, duplicate detection, Windows suffix, preparation order and attestation checks remain.
• Runtime scheduling continues after a failure; clean-start and failed-measurement weight cases remain.

How to verify it

Focused Deno command, passing 31 top-level tests / 202 steps:

deno task test packages/cli/tests/repl-execution.test.ts packages/core/tests/expand.test.ts packages/cli/tests/evaluate-workflow.test.ts scripts/tests/staged-write.test.ts scripts/tests/release-targets.test.ts scripts/tests/runtime-tests.test.ts scripts/tests/test-weights.test.ts scripts/tests/runtime-exclusions.test.ts

Ten temporary controls failed as intended: teardown release, early occurrence settlement, shared subscription, obsolete inputs performing effects before refusal, changed fresh-run output, incomplete replacement payload, swapped macOS artifacts, first filename paired with a later failure status, stale weights and inexact published bytes. Quoted/reordered YAML preserving the pairs passed. Every temporary fault was restored byte-identically before commit.

SR8 explicitly places the existing cause table outside the occurrence, matching a real execution; otherwise direct expansion's fallback scope can join a leaked producer and mask removal of the construct-owned scope.

Architect feedback: PASS for exact commit 39317e4e97281fb867eaf320583302b9b5801a4d against base 62d1086703ec72ca6d3b6e90a596ee716c571237.

Separate local delivery verification passed: deno task lint (warnings, zero errors), deno task check, tsc --noEmit -p tsconfig.node.json, and deno task check:jsr (Success Dry run complete). Focused Node and Bun runs each passed 172 tests across the six portable changed/affected files. Evaluate workflow and release tooling retain their existing runtime exclusions. CI owns the exhaustive runtime corpora; its status was not inspected for the feedback verdict or publication.

Scope

Included

• First-slice assertion and contract-description corrections from Make XMD test assertions, fixtures and failures trustworthy #914.

Intentionally unchanged

• Packaging ownership and stale-output fixtures, diagnostics/navigation, error classification and optional consolidations remain later slices of Make XMD test assertions, fixtures and failures trustworthy #914. This PR does not close that issue.
• Production behavior, dependencies, release configuration and runtime eligibility remain unchanged.

Scope confirmation

• Every changed file supports the purpose described above.
• Unrelated cleanup and formatting changes are excluded.
• Generated or mechanical changes are clearly identified: none.
• The description matches the final diff and test results.

[github-actions]

PR #915: 🧪 Make lifecycle and publication tests observe complete outcomes

9 files, +199 / -60

Scope

✅ PR scope looks good.

Structural

✅ No structural bloat detected.

Slop

✅ Slop indicators look low.

Static Analysis

✅ Oxlint found no issues.

Correctness

No extraneous code patterns detected.