[Tracker] Track nightly and CI failures for the 26.10 development cycle

[csadorf]

This issue tracks nightly and CI failures that affected main during the 26.10 development cycle and release/26.10 after burn-down started:

• [CI] Intermittent failure of cuml.accel test_voting::test_sample_weight[42] test #8384
• [BUG] Dask logistic regression tests deadlock (multi-GPU only) #8398
• [CI] Network xfails cause sklearn examples to miss pass-rate threshold #8410
• [CI] Failing integration/test_rf_classifier.py::test_oob_score in wheel-tests-cuml-nightly-dependencies-versions #8425
• [CI] Failing test_estimator_proxy.py::test_set_callbacks_meta_estimator in wheel-tests-cuml-nightly-dependencies-versions #8426
• [BUG] Failing tests on GB300 #8430
• [CI] RandomForestRegressor sklearn compatibility check is flaky #8457
• [CI] Dask OneHotEncoder CuPy test fails during task serialization on Python 3.14 #8487
• [CI] Tracking RandomForest Failures #8507
• [CI] CD Test Failures #8508
• [CI] SG TSNE Test Failures #8509
• [CI] New test failures on GB300 #8510
• [CI] Nightly scikit-learn tree node dtype breaks RandomForest and IsolationForest conversions #8537
• [CI] plot_outlier_detection_bench fails with a cuDF MixedTypeError #8547
• [CI] IsolationForest as_sklearn tests fail against nightly scikit-learn #8548
• [CI] Docs build cannot fetch cuDF and RMM 26.10 inventories #8557
• [CI] CUDA 13.3 pip devcontainer builds fail during CMake configuration #8559
• [CI] RAPIDS XGBoost packages block 26.12 conda environments #8583
• Change to scikit-learn tree node type breaks treelite.sklearn.export_model dmlc/treelite#696

[betatim]

Fixing dmlc/treelite#696 is probably the answer to this mornings failure in the job that uses the nightly version of scikit-learn