Metal: set_vector_bytes passes nullptr to setBytes for empty vectors

`mx.random.categorical(logits, num_samples=M)` allocates O(N·M) memory

[BUG] ring and jaccl both fail to connect (error 60/65) on 4-node M3 Ultra cluster — IP network verified routable, full mesh confirmed

Strided slice .at(...).add corrupts overlap-add accumulation on MLX 0.31.2

head_dim=256 unfused SDPA makes 100K+ context prefill impractical — real-world evidence for reviving #3293

[BUG] JVPs of power, divmod, and slice_update mishandle partly traced inputs

[BUG] JVP silently returns a zero tangent when differentiating through mx.where

compile: failed first trace leaves a stale cache entry that returns empty-success on retry

[BUG] gemma-4-12b with omlx error loading model

[BUG] All_reduce over >INT_MAX elements hits RuntimeError