Transform tweens are calculated and written linearly on a single thread, likely the main-thread. This may be faster than using PhysicsWorld.TransformTweenMode.Parallel if the majority of the are not split across hierarchies so that they can be written in parallel. To further clarify, if most of the Transform are not interleaved across different hierarchies, this non-parallel (sequential) mode may be faster than PhysicsWorld.TransformTweenMode.Parallel, because it avoids the overhead of splitting and synchronizing work across multiple threads when there is not enough independent hierarchy work to parallelize efficiently.