Review of PR #3200 found a gap in the all-skipped refresh fix imported from #3185: the new EXISTS predicate proves only one processed effect somewhere after the base, not that the claimed target frontier is supported.
Concrete proposed regression: active base1, skipped effect2, request target3 with no effect at3. The empty materializer can persist progress through3; EXISTS sees2 and permits activation. The existing new denial test has no effects at all and misses this case.
Review evidence: #3200 (comment)
Reproduce through the real refresh API, then validate target coverage against the canonical processed observation authority. Preserve legitimate all-skipped completion, frozen source identity, and denial of unsupported advancement; do not treat copied progress/receipt fields as proof of source coverage.
Resolution: Merged via #3141 (which contains #3200). Five real refresh regressions verify exact target-frontier coverage, including skipped/message prefixes and foreign-session target effects; legitimate all-skipped completion is preserved.
Review of PR #3200 found a gap in the all-skipped refresh fix imported from #3185: the new EXISTS predicate proves only one processed effect somewhere after the base, not that the claimed target frontier is supported.
Concrete proposed regression: active base1, skipped effect2, request target3 with no effect at3. The empty materializer can persist progress through3; EXISTS sees2 and permits activation. The existing new denial test has no effects at all and misses this case.
Review evidence: #3200 (comment)
Reproduce through the real refresh API, then validate target coverage against the canonical processed observation authority. Preserve legitimate all-skipped completion, frozen source identity, and denial of unsupported advancement; do not treat copied progress/receipt fields as proof of source coverage.
Resolution: Merged via #3141 (which contains #3200). Five real refresh regressions verify exact target-frontier coverage, including skipped/message prefixes and foreign-session target effects; legitimate all-skipped completion is preserved.