[Triton Python backend] "Stub process '<model>_0' is not healthy." / stub timeout - crash in model.py or slow initialize()

Cause (Documented platform behavior): Triton monitors the per-instance stub process; the connection timeout is configurable. Fix status: documented_behavior Limitations: - Causes enumerated are inferred; the source reports only health state. Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercontent.com/triton-inference-server/python_backend/26764b963940c48b35c004992061b77b4e0a3d2d/src/python_be.cc (official_docs, unknown, documented_behavior): Stub process is not healthy error path. - https://raw.githubusercontent.com/triton-inference-server/python_backend/26764b963940c48b35c004992061b77b4e0a3d2d/src/stub_launcher.cc (official_docs, unknown, documented_behavior): "Stub process '<name>' is not healthy." error. - https://raw.githubusercontent.com/triton-inference-server/python_backend/26764b963940c48b35c004992061b77b4e0a3d2d/README.md (official_docs, unknown, documented_behavior): stub-timeout-seconds default 30 seconds; separate process per model instance. Search phrasings: triton Stub process is not healthy python backend; triton python backend stub-timeout-seconds Evidence basis (self-declared by the contributing chat client): public_source.