Documentation verification summary

On 5 November 2026, 29 focused receiver tests passed with zero failures or skips. A separate local probe passed one eligible baseline and seven blocked delivery cases. Receiver revision: c721c37eb83b011678b174edfb11221c32c9ac19; Node v24.22.1 on macOS. Tests used isolated temporary data and local listeners. No live pilot, owner approval, phone operation or deployment was involved.

Claims and checks

Claim	Specific check or source	Result and scope
Older supported SDK messages remain accepted	`native-upload-contract.test.js`: actual SDK overlay/depth queue passes authorized receiver, canonical replay and exact retries	Passed for retained preview.1 fixtures, not freshly generated build 43 archives
Receiver validates descriptors and preserves malformed indexed evidence	Same file: unsupported contract descriptors and evidence totals fail before writes; malformed archive index preserves source chunks	Passed; not exhaustive maximum-size boundary testing
Compression preserves original identity	`native-upload-encoding.test.js`: upload compression preserves exact envelope bytes and caps expanded input; `native-gateway-compression.test.js`	Passed for bounded decoding and forwarding
Bundles preserve per-message receipts	`native-upload-bundle.test.js`: overlapping regrouped retries; invalid shape/identity/order; partial quota rejection	All three passed, including preservation of already-stored messages
Concurrent approvals preserve withdrawal	`approval-race.test.js`: seven delayed approval/withdrawal, stale asset and stale placement cases	All passed
Withdrawal retains review history and blocks reactivation without approval	`approval-withdrawal.test.js`	Passed with isolated HTTP fixture, ownership and stale-review rejection
Withdrawn artwork can be reviewed again	`placement-review-withdrawal.test.js`	Passed in local review-state fixture
Upload recovery does not renew presentation or enroll unrelated history	`native-recovery.test.js`: expired opted-in upload recovers after withdrawal; historical recording without opt-in rejected	Both passed
Browser durations retain distinct meanings	`measurement.test.js`: version 2 durations distinguish observed, qualified and longest streak; malformed/unknown/gap cases	Eight tests passed
Pause, disable, categories, approvals and review revisions block delivery	Supplemental `delivery-controls.mjs`, using the same `deliveryProblems` function referenced by the server	Baseline plus seven blocked cases passed; not an end-to-end endpoint or hosted test
Native fragments can total qualified time without an impression candidate	Existing build 43 SDK release log: interrupted fragments never combine into an impression	Previously passed; not rerun in this revision

The first suite attempt could not open local listeners. It is retained as an unsuccessful environment-limited run. The final run enabled local listeners and passed. An intermediate launch could not find Node; the final run used the installed executable explicitly. The passing result does not erase those attempts.

Specification reconciliation

The earlier upload prose describes the smaller common SDK subset. The maintained compatibility reference now distinguishes that subset, the pilot sender and receiver admission limits. Sender caps remain stricter than receiver caps.
Earlier introductory marketplace text lists 1.5% screen coverage, two seconds and a 260 ms gap. The inspected current browser policy uses 2%, one second and 250 ms. The measurement and approval specification records the current meanings without rewriting historical receipts.
Earlier assignment-upload prose omits later transport bundling and gzip. The maintained reference separates transport encoding from immutable queued messages and from saved compressed-copy recovery.
The audit plan alone was incomplete evidence. The historical final-verification record maps all 16 findings to implementation commits and regression references. Its integration results were 446 marketplace and 144 SDK checks at their recorded revisions. This review rechecks selected controls; it does not certify every finding against today's code.

Historical source snapshots and specifications are retained for provenance. Use these maintained references for the stated revisions; do not combine limits or policy values from older introductions with newer pilot behavior.

Remaining boundaries

This closes the local documentation traceability work for receiver limits, approvals, withdrawal and measurement definitions. Current hosted revision/state, exhaustive maximum-size uploads, broader device behavior and independent assessment remain outside this verification. See the installation guide for the separately checked SDK package.