How do you run JS on the GPU?
fast large matrix transforms in JS on the GPU
You still seem to be failing to understand that these limitations would only come into play when cross-origin textures are in use and the alternatives are (a) don't allow cross-origin textures, as is the case now, (b) restrictions on what you can do in the shaders or (c) a solution nobody has thought of yet.
In the end you're dealing with a machine you can't change and with usecases you wouldn't want to kill. So I think accepting that you will always be able to use timing attacks is essential, because the very same property that makes timing attacks possible makes useful usecases possible.
There is no need for adding artificial fuzz. Browsers already limit the maximum rate renderAnimationFrame will be called. Therefore disabling finish() will prevent reliable timing of anything that renders in less than the minimum time between calls to renderAnimationFrame. What's left is to deal with renders that take longer than that. For that I suggested aborting shader execution after a set time possibly coupled with improvements to GLSL optimization and asked which approach is least objectionable: today's status quo, CSS filter restrictions or my proposal. The question should be answered from the perspective of the applications mostly likely to want to use cross-origin textures.
You could attempt to add "artificial fuzz" but that's just degrading everybodies performance because: blocking time of render+ random fuzz > blocking time of render
NOTE: This electronic mail message may contain confidential and privileged information from HI Corporation. If you are not the intended recipient, any disclosure, photocopying, distribution or use of the contents of the received information is prohibited. If you have received this e-mail in error, please notify the sender immediately and permanently delete this message and all related copies.