v1.8.0 — Headless sweeps and the measured ceiling

1.8.0 — Headless sweeps and the measured ceiling

• Keep Frame awake on the Overview card renews the headset's own 60-second activity
  extension every 20 seconds over the paired SSH link, like Auto tune's unattended mode, so a
  Frame on a desk keeps streaming. The capacity sweep and Restart Frame stream hold the same
  extension while they run; it ends when switched off or when the app closes.
• The sweep's highest clean target is kept across restarts and saved as a personal preset
  ("Measured ceiling N Mbps …") built from the current setup with only the target changed. The
  card compares it with Valve's 200 Mbps automatic starting rate, and Apply N Mbps now puts
  the measured ceiling on the running stream in one click with automatic bitrate off.
• The sweep measures a marginal or failed step a second time before it counts (twice per sweep
  at most) and confirms the winning step with another pass before saving it as the measured
  ceiling, so one burst of interference neither ends the sweep early nor becomes the recommended
  rate. A failure that does not repeat counts as marginal: the sweep goes on, but that rate is
  not recommended on its own.
  Two-pass steps are marked in the results and the summary names every marginal target. Each
  sweep replaces the measured-ceiling preset the app saved before instead of adding another.
• The card keeps the last 20 sweep results and, once two fall within a week, says whether they
  agreed within a step or ranged, with advice for a link that varies: settle on the lower value,
  or use automatic bitrate starting at the higher one so SteamVR backs off when the link dips.
  For the first 90 seconds after a sweep the card says the live counters still include the
  sweep's highest steps instead of arguing with the result.
• The sweep's start dialog offers Apply the measured ceiling when the sweep finishes (on by
  default, remembered), so one click measures the link and puts the result on the stream. The
  summary now states how much video the winning step actually averaged and calls the result
  lightly tested when that was under 85% of the target, since a static scene never asks the link
  for the full rate.
• Reading steamvr.vrsettings right after a change no longer fails with "being used by another
  process": SteamVR holds the file exclusively for a moment while it saves, and the read now waits
  it out. The sweep's measured-ceiling preset was the first victim.
• The sweep can show Auto tune's moving test scene in the headset while it measures (ticked by
  default when nothing is playing), so a headset on the desk tests the link at the full rate
  instead of a still scene's third of it. Receiver retry requests are now listed per step and
  count against a step only at 60 or more per minute; held frames and the headset loss ratio
  remain the judges, since a full-rate scene logs a few retries per step on a healthy link.
• A sweep that finds no clean step now leads the card (nothing above its best marginal rate is
  worth trying now), retires the saved ceiling instead of leaving an "Apply N Mbps" button for a
  rate the link just refused, and is kept in the history as such.
• Apply N Mbps now is no longer a permanent button: it appears for 20 minutes after a sweep
  whose winning step the encoder actually filled, and older or lightly tested ceilings stay in
  the text and the Measured ceiling preset. Lightly tested results are left out of the history
  range and of the Automatic from N Mbps suggestion.
• Before the first step the sweep and the check wait for ten quiet seconds without a stream
  recovery request (up to 40 s), so a stream that has just restarted is not scored as failing.
• Check this rate (30 s) measures the live target once, as one sweep step with the same
  rules, re-check and test scene, and reports whether that rate is clean today without changing
  or saving anything.
• Automatic from N Mbps appears once the week's clean ceilings disagree by a step or more:
  one click turns automatic bitrate on with SteamVR starting at the highest ceiling measured
  instead of Valve's 200, with the manual target raised so it does not cap the start.