Salesforce: fix the cross-org Lead lookup, and drop the no-op sfdx logout

[vishesh1999gupta]

Two independent fixes to the Salesforce tests. Both verified against the shared test org.

---

1. The cross-org Lead lookup rejected duplicates

The sink tests that verify their record reached the second org intermittently failed at the
very end, with every assertion already green and no error shown:

✨ Display content of topic success-responses, it contains 1 messages
✨ Display content of topic error-responses, it contains 0 messages
   Login with sfdx CLI on the account #2
   Authorized to [email protected]
   Get the Lead created on account #2
🧹 Cleaning up: Lead John_24730 Doe_25713 (both orgs) …
🔥 RESULT: FAILURE

Cause

sfdx data:record:get --target-org "$SALESFORCE_USERNAME_ACCOUNT2" -s Lead \
  -w "FirstName='$LEAD_FIRSTNAME' LastName='$LEAD_LASTNAME' Company=Confluent"

data:record:get requires the where-clause to resolve to exactly one record:

Error (1): FirstName='…' LastName='…' Company=Confluent is not a unique qualifier
for Lead; 2 records were retrieved.

The sink connectors insert — never upsert — so a shared org accumulates Leads, and any
run that dies before its cleanup trap leaves one behind. LEAD_FIRSTNAME is John_$RANDOM
(0–32767), so a later run redrawing that pair fails permanently, even though the record
it just wrote is present and correct. The org used for these tests holds ~500
Company='Confluent' Leads, so orphans accumulate and each poisons its own name pair.

The error was invisible in CI because the command runs under set -e, so the following
cat /tmp/result.log never executed — the run showed only a bare RESULT: FAILURE.

Fix

data:query returns every match and exits 0. The existing grep still fails when the
record is genuinely absent, so the assertion keeps its meaning — it just stops requiring a
uniqueness the tests never guaranteed. || true is added so cat always runs.

Verification

Scenario	data:record:get	data:query
2 matching Leads	exit 1, "not a unique qualifier"	exit 0, both rows, grep passes
0 matching Leads	exit 1	retrieved: 0, grep fails — regression still caught

Full salesforce-sobject-sink.sh run with one duplicate pre-seeded in the second org:

• before — RESULT: FAILURE, is not a unique qualifier for Lead; 2 records were retrieved
• after — RESULT: SUCCESS, /tmp/result.log showing both rows

Applied to all six affected tests, including the proxy variants.

---

2. sfdx force:auth:logout --all in cleanup was a no-op that broke later steps

The cleanup traps ended with a logout, justified by a comment claiming a run "accumulates
sessions against the same user and later logins evict earlier ones". Both halves are wrong:

• No per-user session cap exists. The Developer Edition limit of 5 is Concurrent API
  Request Limits (requests running ≥20s) and returns REQUEST_LIMIT_EXCEEDED rather than
  evicting.
• It makes no network call. In the image these tests use (sfdx-cli 7.209.6,
  plugin-auth 2.8.4, @salesforce/core 4.3.10) it resolves to deleting
  ~/.sfdx/<user>.json; there is no reference to services/oauth2/revoke in either
  package. Sessions lapsed on idle timeout either way.
• It is redundant. The sfdx-cli container is recreated per test — hence
  Config folder does not exists, Creating Folder on every login. The store already starts
  empty.

What it did do is delete credentials later steps need. Querying the second org after a test
consistently failed with No authorization information found for <user> until the account
was re-authenticated.

Verified: salesforce-sobject-sink.sh passes with the call removed, and a query against
the second org afterwards now succeeds with no re-login — previously it always failed.

---

bash -n clean on all changed files.