[barelyworkingcode]

Closes #573

Outside-session e2e tests read a failed probe's snapshot under load. osRig.probe() and probeFrame() now share probeRaw(), which returns only once every machine's lastProbeAt moved and none is unreachable; after a failed probe it moves the clock again, inside the same osBound.

TestOutsideReportForExitedProcessRemovesRow's hook wait missed for the same cause: osSlowProbe's probe() returned a timed-out, unreachable frame, so the test killed the process and reported the hook while the machine was unreachable. Not a relay defect.

Fix attempts: 1. Tokens: about 66k dev, 45k review.

Minor review findings (not fixed): the probe() doc comment does not say an unreachable machine is rejected; the wording in probeRaw's comment could be tighter.

Proof:

• Before: 32 CPU burners, -run 'Outside|Adopt|Ended' -parallel 64: 13 failures with this bug's signature ("list held", "ended rows", "want idle", "no row for"). Same run after: 0 with the signature. Other failures there are frame timeouts from starvation and adopt-door 404s, outside this contract.
• After: the four tests under 64 burners -count=3: ok. Alone: ok. -run 'Outside|Adopt' unloaded: ok.

Review: clean, no blocking findings (Sonnet, round 1)
Red on main: n/a: three of the four tests exist only on the epic branch; red on the epic head bb10fff under forced CPU load (13 signature failures), 0 after the fix
Real-app check: n/a: test-only change under e2e/, no product code moved
Journey: n/a: no devboxverify journey or product behaviour changes

🤖 Generated with Claude Code

[barelyworkingcode]

merge-check: merged. approved, target, contract, evidence, owner, weakening

[barelyworkingcode]

Merged into epic/525-session-adoption: approved, target, contract, evidence, owner and weakening PASS; ci and verified NA.