Original Maelstrom exact select capture: preguest timing blocker
Historical source last updated, 2026-10-24. Runtime and implementation claims below describe that dated source; see current app notes and source for subsequent changes.
The original guest was not started. The earliest missing milestone is a guaranteed host capture after actual result/EAX storage and before any later guest execution. No select arrival, park, completion, result/token/route/thread/ESP/saved return, fdset or timeval observation from the original game exists.
The unchanged accepted hook can bind createOracle(actualInstance, actualInstance.exports) and snapshot(actualInstance.exports) at a synchronous host boundary. That binding alone cannot guarantee the required timing with ordinary run(1000). The exact transformed handle_select calls msc_end after its real EAX store. msc_end stores only diagnostic globals and calls only msc_fail on invalid state; it does not yield, stop the budget or invoke the host. On completion, yield reason/flag remain zero and steps remain unchanged. The production thunk path sets the saved return EIP and branches back into the run loop, which can execute subsequent guest code before JS regains control. The standing runner's immediately-after-run boundary therefore cannot establish the requested before-next-guest guarantee. Bounded fdsets/timeval read there could also postdate subsequent guest writes.
This is a source-backed inability to guarantee the requested timing, not proof that a particular original invocation executed further instructions. No such invocation was run. No production repair, compilation, module change, synthetic test_call_select, observer layering, pacing change, breakpoint or extra run was attempted. No adapter was installed after the temporal prerequisite failed. A completion fence/callback qualified for the unchanged scheduling would be needed before a new coordinator-assigned capture; implementing one is outside this task.