WebGL conformance failure: angle-instanced-arrays-out-of-bounds.html expects INVALID_OPERATION but Chromium returns NO_ERROR

[sarathsi]

We are observing failures in the following WebGL conformance test when running on Chromium-based browsers:

Test: https://registry.khronos.org/webgl/sdk/tests/conformance/extensions/angle-instanced-arrays-out-of-bounds.html
Specifically, some cases in the section: "Test out-of-range instanced attributes" are failing.
FAIL getError expected: INVALID_OPERATION. Was NO_ERROR : after evaluating: ext.drawArraysInstancedANGLE(gl.TRIANGLES, 0, 9, 1)

Attempted Fix
A change was proposed in ANGLE to explicitly generate GL_INVALID_OPERATION via additional CPU-side validation:
https://chromium-review.googlesource.com/c/angle/angle/+/7588515

Feedback from ANGLE Code Owners
The required CPU-side validation to detect these out-of-range instanced attribute accesses is extremely expensive.
When backends ensure robust buffer access and memory safety, returning GL_INVALID_OPERATION may not provide meaningful additional value. Therefore, it was suggested that the WebGL specification might be reconsidered to avoid requiring this error in such cases.

[ShabbyX]

For completeness, the test is marked as failing in many configs in Chromium: https://source.chromium.org/chromium/chromium/src/+/main:content/test/gpu/gpu_tests/test_expectations/webgl_conformance_expectations.txt;l=407;drc=e818e9a078b2bb2821a5f0341a7776500ebac7d1

Also, according to @lexaknyazev:

The WebGL position is as follows: if the platform has robust access, then draw calls that might go OOB do not generate any errors.

So it would be wrong for the tests to expect an error code.

[lexaknyazev]

Yes, the suppressions seem to confirm the test bug.