Autonomous PID 1 reaches both markers and interactive Bash

The one physical comparison produced strict fresh-nonce B then E, followed by an interactive Bash primary prompt. Linux 7.3-rc5, the exact selected kernel arguments, ttyS0 registration/console enable and /bin/sh init entry qualified the records. The passive capture lasted 60.035 seconds, sent zero input, and recorded no protocol errors. Zero attempted input is supported by the reviewed passive controller and absence of later command frames; it is not an independent electrical direction measurement. RX remains NOT_TESTED. No candidate reboot or persistent selection change was requested.

The selected fixed script first tests PID 1 and UID 0, prints B, runs /bin/sleep 5, prints E and executes interactive /bin/sh. B supports the limited builtin tests and reaching the first output. E supports successful return from the preceding B printf and selected sleep. Each marker alone does not prove its own output call returned; the later primary prompt adds progress through the E output and interactive-shell handoff. The five-second argument is verified, but this UART transcript does not timestamp each marker independently. No full /proc identity guard or runtime instruction/IRQ trace was collected.

This establishes userspace execution and sleep completion in this boot, with all optional progress-reporter/trace/tick gates absent. It narrows claims that all mainline userspace or timed waits universally stall. It does not identify why ordinary init or the earlier enabled reporter failed to progress, prove serial RX, or show the normal root/panel/glass. Ordinary init was not attempted in this comparison; task 5b.5 remains UNVERIFIED.

Root operated ~/tmp/k230-mainline-probe-integration, branch integrate/mainline-probe-path, frozen at c0c8e50c47abfb165d3c4b4c0b51b2653fb48e91. The preceding nohz-off capture first passed NEW protected normal recovery. A fresh byte-identical host qualifier then passed against that report and the same p2/24h/0l4 artifacts, native receipt, archive, loader and Bash exports; its executed safe receipt is preserved. The runtime controller independently rechecked protected normal state and all five load/CRC guards before its exact volatile printenv check and sole candidate bootm. No new build, transfer, flash or readback was performed.

python3 tools/mainline-drm-initrd-shell-trial.py --mode minimal \
  --same-image-shell-pid1 --autonomous-bash-pid1 \
  --bundle /nix/store/p2kdar89q7dhwdajjy0mrhmxxzcgsalw-k230-mainline-drm-trial-boot-files \
  --manifest PRIVATE_MANIFEST --normal-report PRIVATE_NORMAL \
  --log PRIVATE_LOG --result PRIVATE_RESULT

Result preserves timestamps, code/artifact identities, limited observations and private transcript hashes. The public fixed records replace only the runtime nonce with the explicitly public native-fixture placeholder. They are a redacted display copy; strict parsing was performed against the private actual nonce and raw wire. Raw UART, boot IDs, runtime nonce and protected reports remain private.

The controller exited 2 because protected normal return was not observed; diagnostic_ok is true. The diagnostic completed, with independent recovery still required when capture ended. A subsequent NEW user reset passed protected normal recovery: distinct boot, exact system/profile/ kernel/init, eight matching boot hashes, three active services and registration absence. The reset used before this comparison was not reused. This follow-up is physical UART evidence; no camera or real-glass proof was obtained for it.

Independent read-only review PASS: five strict load counts/CRCs, exact volatile arguments and final boot command, complete fresh B/E ordering and later prompt, and pure chunked-parser replay agree with the saved result. The preceding NEW protected normal recovery was also reviewed against its unique raw postflight receipt. No reviewer opened UART or performed a build.