[CaelmBleidd]

Calls previously measured members independently and mixed native coverage with symbolic-result agreement. Measure exact executable original TypeScript statements of one class under a shared 30- or 120-second budget, retaining input candidates when symbolic execution reaches an unsupported call or an execution limit.

The frozen class surface schedules the constructor, every member and ordered two-member scenarios on one receiver. A single machine retains scenario queues with constant-time fairness, a 250 ms quantum, Yices and class-only symbolic novelty. The original AST denominator stays fixed across profiles and budgets; wrappers, foreign classes, model bodies and type-only declarations stay outside coverage.

Primary coverage requires complete, agreeing original and independently instrumented native replays with checked target-class statement IDs. Symbolic returned-value/state consistency is a separate strict diagnostic. Candidate origin participates in deduplication, so a stopped prefix cannot hide a later terminal observation with the same inputs. In-search candidates retain their actual emission times; final queue extraction contributes only to the final budget checkpoint. Candidate caps, extraction failures and overruns remain visible.

Class measurement connects TsStoppedStatesObserver to the caller's candidate consumer. It snapshots feasible unknown-call prefixes before pruning, observes reachable runtime-bound inputs, and collects queued inputs after successful analysis while the machine remains open. The shared UMachine.run, observer and options contracts come from main unchanged. Collection performs no additional execution steps or SAT checks and preserves stop metadata.

JacoDB remains pinned to 37327563ab from UnitTestBot/jacodb#437 for constructor parameter-property binding. Schema 2 binds coverage and stopped-input policies; source closure, harness, domains, models, frontend and clean tool revision are bound into each cell. Source/instrumentation corruption, foreign hits and resume drift invalidate the acquisition attempt.

Rebased onto main at 26f00125a3eb7ebbb1fbc8fcff8e1813f2e3926b. Already-merged engine changes are excluded from this PR; the additional observer migration touches three Calls files.

Local validation of the rebased code: 64 Kotlin class/coverage tests, 41 Node source-inspection tests and 22 Python driver tests passed, with no failures or skips. Calls Detekt, whitespace checks and installDist passed. Broader suites remain for CI. These checks establish fixture behavior and packaging; they are not a rerun of the scientific campaign.

Historical evidence at 421dd926ee39312ec678bfc23be5390066f5a398: one unchanged HashMap, seed 17, a requested 30s in each profile, 91 scenarios and 66 original-class statements. All 475 candidate pairs passed complete agreeing original/instrumented replay without extraction, replay or native infrastructure failures.

Profile	Primary native /66	Strict secondary /66	Terminal	In-search prefix	Post-stop frontier
EMPTY_STOP	57	0	0	91	0
EMPTY_FRESH	43	0	10	0	0
FROZEN_STOP	58	20	5	0	182
FROZEN_FRESH	58	20	5	0	182

Those measurements are a failure reproduction on one class. The sealed v7 campaign covers three unique classes and retains its historical strict metric; neither data set establishes ecosystem representativeness. The historical results remain tied to their original revision and were not recomputed by this rebase.