# Auditing my OWN 1 697 with the port's rule -- and the first correction failed
# its own control. 2026-08-30.
#
fallback fires                         2305
  of those, rest alpha > 0             1697
    element's LAST keyframe alpha > 0  347   <- ends visible; resting visible is CORRECT
    element's LAST keyframe alpha = 0  1350   <- fades out; a visible rest is a transient's peak
  rest alpha == the element's MAX      1457

CONTROL on the split itself — is 'ends at a=0' near-universal?
  all elements with >= 2 keyframes     13991
  of those, last keyframe alpha = 0    12278  (87.8 %)

--- END (if this line is missing, the run did not finish) ---
#
# THE STORY, SAID OUT LOUD:
#
# 1. '1 697 fallback fires return a visible pose' was published as if the
#    number were a defect count. It is not. An element that genuinely ends
#    visible and stays visible SHOULD rest visible; the fallback being the
#    path that got there is not an error.
#
# 2. FIRST CORRECTION, and it looked clean: split by whether the element's
#    LAST keyframe is visible. 347 end visible (correct), 1 350 fade out
#    (a transient's peak). Plausible, arithmetic fine.
#
# 3. 🔴 IT FAILED ITS CONTROL. The port agent had just found that a SCREEN's
#    exit ramp drives every element to a=0 at the end -- which is why its own
#    census called ptmsg, the permanent footer, 'a 2-unit flash'. Measured
#    here: 12 278 of 13 991 elements (87.8 %) end at alpha 0. So 'ends at
#    a=0' is near-universal and says almost nothing about being a transient.
#    The 1 350 is not a transient count and is NOT published as one.
#
# 4. ✅ WHAT SURVIVES, and it needs no such split:
#
#      2 305  elements where the dwell fallback decides
#      1 457  of those rest at the element's MAXIMUM alpha
#
#    The fallback runs only when NO two adjacent poses are equal -- i.e.
#    only when no pose is held. So every pose it can return is un-held by
#    construction, and 1 457 times it returns the BRIGHTEST un-held pose.
#    That is the defect shape, stated without needing to know where the
#    element's visibility ends.
#
# I ran the control only because the port had just been bitten by the exit
# ramp. Without that message the 1 350 would have shipped.