Exercise 3: Testing ADR-007 Against Chapter 9's Own Four-Question Standard — Possible Solution
====================================================================

THE FOUR QUESTIONS, MAPPED TO THIS CHAPTER'S OWN ADR-007
------------------------------
    Why not two separate implementations? -> answerable: True
      (Context section: "This was rejected: Chapter 8 verified that two
      independently-implemented copies ... can silently diverge")

    What was actually decided? -> answerable: True
      (Decision section: "will hold a reference to the same PricingEngine
      instance ... constructed once")

    What do we lose/gain by doing this? -> answerable: True
      (Consequences section: "guaranteed to match ... a coordination cost")

    Which alternative was rejected and why? -> answerable: True
      (Context section: "This was rejected: Chapter 8 verified...")

    TOTAL: 4 of 4

A METHODOLOGICAL NOTE, FOUND WHILE VERIFYING
------------------------------
A first attempt at this check used a naive substring search for the
exact phrase "this was rejected" and got a FALSE result for that
question - not because the ADR was missing the information, but because
the phrase happened to be split across a line wrap in the source text
("This was\nrejected:"), so the literal substring never appeared on one
line. Normalizing whitespace (collapsing all runs of spaces/newlines
into single spaces) before searching - the same way a human reader
would naturally read the wrapped text as one continuous sentence -
fixed the check and correctly returned True. This is an honest caveat
about the four-question test itself: naive keyword search can produce
a false negative purely from formatting, not from the ADR's own actual
content, so a real answerability check should normalize whitespace
before searching, not search the raw text verbatim.

WHY ADR-007 SCORES A FULL 4 OF 4
------------------------------
ADR-007 was deliberately written to include all three of Chapter 9's
own required sections (Context, Decision, Consequences) in full, rather
than a shortened version - directly following Chapter 9's own GOOD_ADR
as a template rather than BAD_ADR's minimal one-line Decision-only
shape. The result confirms that following Chapter 9's own established
structure, section by section, reliably produces an ADR that passes its
own test - the structure isn't decorative, it's what makes the
questions answerable in the first place.

WHY THIS WORKS AS AN ANSWER
------------------------------
Each of Chapter 9's own four questions is checked against ADR-007's
actual text with the specific answering section identified, not just a
pass/fail total, and a genuine methodological issue found during
verification (the line-wrap false negative) is reported honestly rather
than silently corrected without comment.