Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@tests/dbz-twin/rac/fuzz-test.sh`:
- Around line 431-445: Update the script header printed by action_help() to
include the new lob-logminer component and to reflect the correct workflow
ordering and RAC cleanup instruction: locate the action_help() function (or the
top-of-file help comment block it prints) and add "lob-logminer" to the
Components list, reorder the example workflow to show "down" before "up", and
add the note "Always run ./fuzz-test.sh down before ./fuzz-test.sh up in RAC
fuzz tests to ensure a clean environment" so the output of ./fuzz-test.sh help
matches the usage text updated in the logs case.

In `@tests/dbz-twin/rac/kafka-consumer.py`:
- Around line 147-159: The loop that checks Kafka topics should wait for all
entries in all_topics (LM_TOPIC, OLR_TOPIC, OLR_LOB_TOPIC) before proceeding;
currently it breaks as soon as LM_TOPIC or OLR_TOPIC exists which can mask an
implicitly created OLR_LOB_TOPIC on subscribe. Change the condition in the
polling loop that examines consumer.topics() (the block using topics =
consumer.topics()) to require every topic in all_topics to be present (e.g.,
check that set(all_topics).issubset(topics) or iterate and confirm each), only
then break and call consumer.subscribe(all_topics) and consumer.poll().

In `@tests/dbz-twin/rac/perf/docker-compose.yaml`:
- Line 47: The JDBC URL string "//${VM_HOST}:1521/ORCLPDB" silently becomes
malformed when VM_HOST is unset; update the interpolation to the required
expansion form so Compose fails fast—replace "${VM_HOST}" with a required
expansion like "${VM_HOST:?VM_HOST is not set}" in that JDBC URL entry to force
an error if VM_HOST is missing.

In `@tests/dbz-twin/rac/validator.py`:
- Line 248: The current LOB detection uses direct membership check ("is_lob =
event_table in LOB_TABLES") which fails for schema-qualified or case-variant
names; update the check to normalize event_table by stripping any schema
qualifier and surrounding quotes/backticks and then compare in a normalized case
(e.g., upper() or lower()) against a similarly normalized LOB_TABLES set. Modify
the logic around the is_lob assignment in validator.py so it derives a plain
table name (split on '.' and remove quotes) and performs case-normalized
membership against LOB_TABLES to correctly detect LOB tables regardless of
schema qualification or case differences.

In `@tests/environments/rac/.env`:
- Around line 1-2: Update the comment for VM_HOST in the rac environment file to
explicitly state that VM_HOST must be initialized by sourcing the vm-env.sh
script (i.e., callers must run or source vm-env.sh before invoking RAC flows),
and mention that rac.sh expects VM_HOST to be set or it will exit (see rac.sh
behavior around its VM_HOST check). Keep the note next to the existing DB_CONN
comment and use the exact symbol VM_HOST and filenames vm-env.sh and rac.sh so
callers know which script to run and which script relies on the variable.

In `@tests/sql/scripts/drivers/rac.sh`:
- Line 39: The strict fail-fast assignment VM_HOST="${VM_HOST:?VM_HOST is
required — source tests/environments/rac/vm-env.sh}" causes the RAC driver to
hard-fail because generate.sh doesn't source tests/environments/rac/vm-env.sh
first; change the driver to not use the "?" expansion (e.g.,
VM_HOST="${VM_HOST:-}") and instead add a guarded attempt to source vm-env.sh
(if present) or bail gracefully only when RAC execution truly requires VM_HOST
later; update the VM_HOST handling in tests/sql/scripts/drivers/rac.sh so it
either sources vm-env.sh when VM_HOST is empty or defers the fatal error to a
later, explicit check that runs inside the RAC execution path.

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@tests/dbz-twin/rac/fuzz-test.sh`:
- Around line 216-225: The readiness loop repeatedly calls docker logs for each
container which rereads the full logs on every iteration; modify the three
docker logs invocations that set LM_OK, OLR_OK, and LOB_LM_OK to include a
bounded tail (e.g., --tail N) or since (e.g., --since "10s") flag so each grep
only scans recent output instead of the entire log; update the commands that
check "Starting streaming" / "streaming client started" in the loop (the lines
that set LM_OK, OLR_OK, LOB_LM_OK) to use --tail or --since to limit work and
improve performance.