A 35-second audio clip killed the vLLM engine for every request that followed.

PR #57769 fixes offline LLM.generate with Whisper crashing on clips over 30 seconds with a tensor size mismatch, after which every later request returns EngineDeadError. The chain: an earlier PR passed truncation=False to HF processors so placeholder text wouldn't be cut, WhisperProcessor forwards that to WhisperFeatureExtractor, which stops truncating audio to its 30-second window, so a 35-second clip becomes 3500 frames and 1750 post-conv positions against a 1500-position encoder. Server transcription endpoints chunk to 30 seconds first and are unaffected, so this only bit offline users. v0.26.0 handled the same input.