Is patching WebGL strings pointless if rendering is still SwiftShader? (F5/Shape, ~40% blocks)

[m9181508445-ux]

Hey, hoping someone here has hit this.

I run a small notifier that checks appointment availability on a Spanish government portal (public .gob.es site) and pings me on Telegram. Detection only, nothing fancy.

Roughly 40% of my checks come back blocked by their WAF (looks like F5/Shape — I get TSPD cookies), and I've been running for weeks without catching anything.

Setup: real Chrome under Xvfb on a Linux VPS, driven over CDP, going out through a Spanish mobile proxy.

Two things I already ruled out with actual tests:

• Ran the same curl_cffi request through the same proxy from a normal laptop and from the VPS — identical responses, byte for byte, zero blocks. So the datacenter itself isn't the problem.
• Not headless, no --headless flag at all.

What I did find is that my Chrome reports platform Linux x86_64 and WebGL "ANGLE (Google, Vulkan 1.3.0 (SwiftShader))" — software rendering, screams "server with no GPU". I patch the UNMASKED_VENDOR/RENDERER strings to look like a normal Windows box, but the actual rendering is still SwiftShader.

So: is patching just the WebGL strings basically pointless against a serious antibot that compares rendered output? Does Botasaurus handle this differently, and is a real GPU on the box the only proper fix?