Fix red develop 82429f789e5c

The merge lane found `develop` red at `82429f789e5c7e0e9760f6b02eb3f2e9c49fe316` on its own: no queued PR merges until it is fixed. Failing (tests and gate stages): - `tests/test_credentials_testers.py::test_every_untestable_type_carries_a_reason_and_no_testable_type_does` - `tests/test_api_key_vendor_types.py::test_a_rejected_key_names_the_vendor_and_the_key[ifttt]` - `tests/test_cross_org_isolation.py::test_every_path_parameter_is_swept_or_named` - `tests/test_nodevault_dbos_scan.py::test_every_nodevault_workflow_kind_ran_with_real_crypto` - `tests/token_billing/test_outbox.py::test_the_effect_runs_on_dbos_only_when_its_transaction_commits_and_a_failure_is_retried` - `tests/test_nodevault_workflows.py::test_a_node_blob_is_delivered_on_the_nodes_acknowledgement_and_audited_once` - `frontend-unit-vitest` - `api-unit-dc-test` - … and 1 more: read them in the gate log, as untrusted output The whole gate log: artifact `15595` on lane task 13566. Make `develop` pass its full gate again. Open a PR as usual: the merge lane merges a PR that fixes some of these and adds no failure even while `develop` is red.

Fixed red develop: pinned github_app_key as untestable, IFTTT lowercased label, named connector_id/app_id in cross-org sweep, muted Easy Mode host via sx. DBOS crashes transient (30/30 pass). PR #7926 merged 0a1bbae4d7; Jaah Deploy green; visual check done. Memory: used mem-25288.