A large share of the profitable kernel fusions in a diffusion serving stack change the output only at the level of bf16 rounding order. Folding the FFN tanh-GELU into the up-projection GEMM as a cuBLASLt epilogue is the canonical example: mathematically the same computation, a different order of operations, and therefore a different last bit.
SGLang's diffusion CI compares generated images bitwise against the diffusers ground truth. So none of these fusions could ever land on the default path. This is not hypothetical — an ungated FLUX GELU fusion was merged as #28166 and reverted in #28708 for exactly this reason.
The usual next move is to argue about the tolerance in CI. That trade is bad in both directions: loosen it and the guarantee everyone relies on is gone, keep it and a whole class of real speedups is unshippable.
The tier
#33453 adds a per-request --quality {lossless, high} tier:
lossless, the default, stays byte-identical to the reference implementation. The ground-truth CI stays green and nobody's guarantee changed.highopts into validated fast paths whose deviation is bounded and visually indistinguishable.
"Visually indistinguishable" is a claim, so it has a gate: a same-seed full-image PSNR above 25 dB is required, and 34–58 dB was measured across every PR shipped under the tier. Side-by-side samples are in each one.
The general lesson is not about diffusion. When correctness is a spectrum and the CI has to be a boolean, the fix is usually to make the request say which end of the spectrum it wants — rather than to move the boolean.