Stabilize nightly testing

[ramakrishnap-nv]

Make the nightly pipeline reliably green across cpp / python / thirdparty / notebook jobs, with alerting on failures.

Already landed (post-26.06):

• Nightly branch bumped to 26.08 (ci: update nightly run branch from release/26.06 to release/26.08 #1424)
• grpc server GPU-memory leak fixed + tests run in parallel (ci(test): fix grpc server GPU-memory leak and run cuopt tests in parallel #1512)
• Multi-GPU (NCCL) cpp test job on 2-GPU runner (ci: add multi-GPU (NCCL) cpp test job on 2-GPU runner #1519)
• pytest-xdist enabled for Python tests (ci(test): enable pytest-xdist for cuopt Python tests #1486)
• nvcc 13.0.3 ICE workaround for sliding_window.cu (fix(routing): lower opt level for sliding_window.cu to avoid nvcc 13.0.3 ICE #1452)
• Unbound CONDA_PREFIX fix (ci: fix unbound CONDA_PREFIX in test_doc_examples.sh #1462)

Remaining:

• Consistent green across all job types; dataset-download resilience
• Close out notebook + third-party sub-issues

Owner: Ram · Size: — · Release: 26.08

[ramakrishnap-nv]

Related open PR:

• tests: retire correctness/regression tests covered by C++ gtests #1461 — tests: retire correctness/regression tests covered by C++ gtests