Exercise 2: DynamoDB's Specific Role, and What Breaks Without It — Possible Solution
====================================================================

DynamoDB's role in the S3+DynamoDB backend pattern is exclusively
LOCKING -- it holds no copy of the state file's actual contents at all
(that's entirely S3's job). Before any operation that would write to
state, Terraform attempts a conditional write to a lock record in the
DynamoDB table; that write only succeeds if no lock record currently
exists. If it succeeds, Terraform proceeds with its operation, holding
the "lock." If it fails (because a lock record is already present),
Terraform blocks and reports the error shown in the chapter, rather
than proceeding.

Storing state remotely in S3, on its own, ALREADY solves the "no
shared record" half of the local-state problem -- every teammate now
reads and writes the exact same object instead of separate local
files. But S3 alone provides no mechanism preventing two concurrent
writes to that same object from racing each other -- S3 will simply
accept whichever PUT request arrives, exactly the same "last write
silently wins" failure mode from Exercise 1, just now happening to a
shared remote object instead of a shared local file. Storing state
remotely changes WHERE the race condition can happen, but does nothing
by itself to prevent the race condition from happening at all.

Without DynamoDB (or an equivalent locking mechanism), two team
members' applies against the same S3-hosted state could still both
read the same state, compute independent plans, and have one's
write silently clobber the other's -- the exact scenario Exercise 1
described, still fully possible even with "shared remote state" if
locking isn't also part of the setup.

WHY THIS WORKS AS AN ANSWER
------------------------------
This correctly separates the two DIFFERENT problems remote state alone
vs. locking alone each solve (shared record vs. safe concurrent
writes), rather than treating "remote backend" as one single fix, and
explains precisely why S3 by itself doesn't prevent the same race
condition Exercise 1 described.