🚗 Element Engine Health

Mechanic's narrative

State of the engine, 2026-09-09 (~129,500 mi, 249 sessions). The motor remains fundamentally healthy: fuel trims sit within normal bounds, closed loop arrives on time, EVAP purge duty-cycles normally, and the electronic throttle tracks the pedal cleanly. The 9/9 morning drive (9-9-26-1, 17 min, S8) is the first genuinely good idle reading since July, and it changes the read on the idle thread.

1. Idle vacuum — the first exonerating datapoint (H2). Warm-idle MAP came in at 10.95 inHg, inside the healthy 9–11 band and the best value since 7/23. For context the Aug 29 – Sep 7 window ran a median of 12.07 with a worst case of 14.39. The detail matters more than the single number: the idle window is one continuous 52-second decay at the end of the drive, RPM falling 920 → 675 while MAP improved 10.6 → 9.2. That is an engine settling normally after shutdown-approach, not a hunting idle. The session's idle_rpm_std of 75.5 (a WATCH-shaped number) is an artifact of that decay slope plus a frozen forward-filled tail — the last 10 s of the 1 Hz grid repeat a single sample (104 raw rows, 2 unique RPM values). Excluding the frozen tail drops the spread to 61; detrending it drops it to 34, i.e. normal. Read this as the A/C-off clean-idle test finally landing, and landing on the "exonerate" side of the 10.5 / >12 threshold that has been posted for a month. It is one session, so it does not close H2 by itself — but it is the first evidence that the August–September vacuum loss is not a fixed mechanical defect. A deliberate 5-minute warm idle is still the confirming test.

2. The cat thread is unchanged — still eroding, still unaddressed (H1/H6). Crossings came in at 9.39/min, which reads high against the 5.43 all-time median but sits well under the historical p90 of 15.44, and rests on only 179 s of cruise. Short cruise windows scatter widely, so this is consistent with the Aug 29 – Sep 7 erosion (median 5.47, max 10.54) rather than an escalation of it. Nothing here revises the picture: no service event is logged after 7/14, so the vapor source is still presumed live and still reloading the cat. The window to act before another full poisoning cycle is still open.

Next data, in priority order: (1) pull a fresh diag report — still nothing since 7/15, now nearly two months stale; with crossings elevated the live question is whether P0420 has re-armed. (2) Repeat the good idle deliberately: one 5-min warm idle, A/C off, at temp, held rather than decaying. 9/9 hints the answer is benign; one clean sustained window would convert that hint into a closed hypothesis and take the valve adjustment off the table. (3) Ask the owner whether anything was done around 7/20 — under the current reading the answer is "no", which closes hypothesis (a) for good.

Ignore-list: lambda_cl_std watches are driven by S8's slower per-PID polling (15 PIDs vs 10) plus short sessions, not real mixture instability — read them alongside stft_std, which is normal. Idle-spread numbers from sessions whose idle window is a single shutdown decay (like 9-9-26-1) should be detrended before they are believed. Warm restarts are excluded from the closed-loop metric programmatically (WARM_RESTART_GAP_MIN in pipeline/baselines.py), so warm relights no longer false-alarm; genuine cold starts hold ~18–32 s and the cold-start verdict stays OK.