Exercise 1: Writes to the Non-Isolated Node During a Partition — Possible Solution
====================================================================

THE TEST
------------------------------
Using this chapter's own CPSystem, with the partition active
(partitioned = True):

    cp.write(node_a, 'stock', 45)   # the non-isolated ("primary") node
    cp.write(node_b, 'stock', 30)   # the isolated node - this chapter's own original test

RESULTS
------------------------------
    write to node_a during partition succeeded, node_a.data: {'stock': 45}
    write to node_b correctly rejected: Write rejected: partitioned, cannot guarantee consistency

Writing to node_a succeeded normally, updating node_a.data correctly.
Writing to node_b was rejected with the identical ConnectionError this
chapter's own text already verified.

WHAT THIS CONFIRMS ABOUT WHICH NODE CP ACTUALLY SACRIFICES
------------------------------
CPSystem.write() only checks `node is self.node_b` before deciding
whether to reject a write - meaning it isn't sacrificing availability
for the SYSTEM as a whole during a partition, only for whichever side
is specifically treated as "isolated" (node_b, in this chapter's own
setup). node_a keeps functioning normally throughout the entire
partition - it never loses availability at all. This matches the real-
world shape of CP systems: a partition typically leaves one side (the
side that can still reach a quorum, or is designated primary) fully
functional, while the OTHER side - the one that can't be trusted to
stay consistent - is the one that goes unavailable.

WHY THIS MATTERS FOR UNDERSTANDING "SACRIFICES AVAILABILITY"
------------------------------
It would be easy to read "CP sacrifices Availability" as meaning the
whole system stops responding during a partition. This exercise
confirms that's not accurate for this chapter's own two-node design -
availability is sacrificed specifically for clients that happen to be
talking to the isolated side. Clients reaching node_a during the exact
same partition see a fully functional, fully available system the
entire time.

WHY THIS WORKS AS AN ANSWER
------------------------------
Both nodes are tested under the identical partition using this
chapter's own unmodified CPSystem, and the asymmetric result (node_a
succeeds, node_b fails) is used to correct a plausible
misunderstanding of what "sacrifices availability" means in practice.