Controlling audio by headset, background processing, serverless data syncing, register capabilities to make the app auto-open for a specific intent, direct connection to the hardware capabilities, etc, etc.

Note that it's not impossible for these things to exist for the web, but by virtue of how these products and standards are developed, web standards will never catch up unless Apple and Google somehow decided that web standards were the way forward and to nerf their native SDK development efforts.

Thanks for your answer. I was asking mainly about performance related capabilities. What can native SDK's do that SIMD, shared memory multi-threading, and GL API's won't address?