The implementation of SharedContextRateLimiter uses OpenGL occlusion queries as fences for rate-limiting GPU-accelerated canvas animations.  This works well on platforms that support the required OpenGL extensions, but on some platforms, notably linux+nvidia, SharedContextRateLimiter falls back to using issueing glFinish every two frames.  Though this does guarantee low latency, it has an unnecessary negative impact on performance and it causes frame rate jank on animations that cannot maintain 60fps.