Exercise 2: Highest-Value-Wins Reconciliation — Possible Solution
====================================================================

THE NEW STRATEGY
------------------------------
    winner_node = node_a if node_a.data['stock'] > node_b.data['stock'] else node_b

Instead of comparing timestamps (this chapter's own last-write-wins),
this strategy picks whichever node's own recorded value is
numerically larger.

RESULT
------------------------------
    highest-value-wins picks: A - final stock = 45
    true combined value: 37
    discrepancy: 8

Since node_a's value (45) is larger than node_b's (42), highest-value-
wins picks node_a - discarding node_b's own recorded sale entirely,
the mirror image of what last-write-wins did (which discarded node_a's
sale).

COMPARING TO THIS CHAPTER'S OWN LAST-WRITE-WINS RESULT
------------------------------
    last-write-wins:    final = 42, discrepancy = 5
    highest-value-wins: final = 45, discrepancy = 8

Highest-value-wins is actually WORSE here - a bigger discrepancy from
the true combined value (8 vs. 5), not better. This is not a
coincidence specific to a badly-chosen strategy; it's a structural
result: for a value that only ever DECREASES from real events (stock
being sold), "highest value" always picks the write representing the
FEWEST real sales - systematically the worst possible choice for this
specific kind of data, since it discards whichever conflicting update
recorded the MOST real-world change.

WHY THIS CONFIRMS RECONCILIATION STRATEGIES AREN'T INTERCHANGEABLE
------------------------------
This chapter's own text already showed one strategy (last-write-wins)
losing real information. This exercise confirms a second, plausible-
sounding strategy loses MORE information for this specific kind of
data - proving there's no universally "safe" reconciliation choice.
The right strategy depends entirely on what the conflicting values
actually represent: for a monotonically-decreasing quantity like stock
depleted by sales, neither strategy tested here is actually correct -
the genuinely correct reconciliation would need to UNDERSTAND that
both writes represent real, additive decreases (a sum, not a pick),
which neither last-write-wins nor highest-value-wins attempts to do.

WHY THIS WORKS AS AN ANSWER
------------------------------
The new strategy is implemented and tested against this chapter's own
exact scenario, its result is directly compared against this chapter's
own last-write-wins figure rather than reported in isolation, and the
counter-intuitive finding (this strategy performs worse, not better)
is explained by tracing the specific reason it systematically picks
the wrong node for this kind of decreasing-value data.