test(server): match V2 tool output redaction in retention coverage

[saphid]

The V2 tool-output redaction change leaves the foundation retention test expecting the old { truncated: true } output preview. Test Server 1 fails on current V2 integrations for #10024 and #10025 because replay and live payloads now intentionally omit that output.

Update the assertion to require omission. The test still checks the transport size limit and verifies that both the stored event and full detail projection retain the original 9 MB output. Production behavior is unchanged.

Validation on V2 base 7dd9873ed: 56 focused foundation/wire-projection tests pass, server typecheck passes, and targeted lint/format/diff checks pass. The stale assertion is reproduced in both linked PRs' Test Server 1 failures. Independent Claude Opus 5 high review was unavailable during this audit: the direct launcher exited 1 with expired OAuth before inference.

Prepared with GPT-6 in the Codex/T3 harness.

[macroscopeapp]

Approvability

Verdict: Approved at 2088c00

Macroscope's review found this PR approvable — This is a self-contained, one-line test expectation update in a test-only file. It aligns replay/live transport assertions with existing output redaction while continuing to verify that full output remains retained in persistence and projections.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

^{You can add or adjust custom eligibility rules. Learn more.}

[juliusmarminge]

Closing as already covered on v2 by commit 8207e55. FoundationPersistence.test.ts now asserts that the transported dynamic-tool payload omits output, while retaining the full persisted-output checks. The exact one-line patch reverse-applies to the current branch, and the replacement commit is already on the remote v2 branch.