SGLang 0.5.22 repairs a hybrid-model state path

The release fixes hybrid linear models resuming from the wrong Mamba state when speculative decoding and structured output are combined. It also requires Python 3.11 or newer. Check the serving image, then rerun structured-output acceptance cases on the affected configuration before rollout. Vendor throughput headlines are separate from this correctness fix. ⁵

Relevant to: SGLang operators combining hybrid models, speculation and constrained output.