Lane CR named all 27 template hooks off, passed check_shim_configs.py, and the
shim installed six detours: the M0 asm stub and the FPU-force module's four
sampling detours, on by default under keys that do not start with hook. The
checker was exhaustive over the wrong list. It now requires fpu.sample_turn and
fpu.sample_ticks off in an exhaustive config and names the M0 stub in its OK
line. All five exhaustive configs amended; their past measurements stand because
each was proved byte-neutral against a control.
Lane BP found both defects while copying cbpin for a determinism probe, and its own
first draft reproduced the comment wrap -- which is what a trap looks like. The
header ran on past the comment and left a stray aiseed.values ahead of the real one,
and six registered hooks went unnamed, so under hooks=trace the config ran six more
detours than it claimed.
The check enforces the two syntax defects everywhere and exhaustiveness only where a
template opts in with '# exhaustive'. hooks=trace legitimately means 'trace all but
these', so a global exhaustiveness rule would be a rule about a style nobody agreed
to; it matters only for a minimal-hook determinism probe.
