Hi,
I don't think we can really know that without trying. FWIW, we have a tracking issue for this at
anglebug.com/42264241, you could use to discuss this in a more centralized way.
VK_KHR_descriptor_update_template is promoted to Vulkan 1.1, which is ANGLE's minimum required version. It's probably not too hard to set it up in ANGLE, and measure the benefits. You are welcome to give that a try.
W.r.t VK_EXT_descriptor_buffer, it's not an extension some mobile vendors can support, but using it in ANGLE could still be beneficial on those that do (and it can get us prepared for some other future stuff coming in Vulkan). Does it need compiler changes too? Either way, my suggestion would be to use descriptor update templates right away, and then look at descriptor buffer.
To answer the ROI question directly, yes I believe both should be beneficial.