Exercise 2: A Better, Still-Inconsistent Team — Possible Solution
====================================================================

SETUP
------------------------------
    reason_recorded probability raised from 0.15 to 0.40

RESULT
------------------------------
    Reason-recorded probability 0.40: 13/20 shipped features have no
    recorded reason they exist

    Chapter's own 0.15 result: 16/20 untraceable
    Lightweight-process result: 0/20 untraceable

COMPARISON
------------------------------
Raising the probability from 0.15 to 0.40 (roughly 2.7x better habits
around writing commit messages) only reduced the untraceable count
from 16/20 to 13/20 - a modest improvement, still leaving 65% of
shipped work with no recorded justification. Even a team that's
meaningfully better than the chapter's own baseline at documenting
their work informally still falls far short of the lightweight
process's guaranteed 0/20.

WHY THIS WORKS AS AN ANSWER
------------------------------
This makes an important point the single 0.15-probability data point
in the chapter couldn't make alone: the problem isn't really "this
particular team is bad at writing things down" - it's that relying on
individual discipline, even meaningfully improved individual
discipline, is structurally worse than a process that simply requires
the information before work starts. No amount of team members trying
harder to remember to comment their commits reaches the lightweight
process's 0% failure rate, because "trying harder" is still
probabilistic, while "the backlog entry requires a reason field" is
not.