[AMD] Dynamic local resource assignment based on thread ID is incorrect

Consider this shader:

RWByteAddressBuffer In : register(u0);
RWByteAddressBuffer OutArr[] : register(u1);

[numthreads(2,1,1)]
void main(uint3 Tid : SV_DispatchThreadID) {
    RWByteAddressBuffer Out;
    if (Tid.x == 0) {
        Out = OutArr[NonUniformResourceIndex(0)];
    } else {
        Out = OutArr[NonUniformResourceIndex(1)];
    }
    Out.Store(0, In.Load(0) + Tid.x);
}

Given that In is a resource with its first element = 42, we would expect OutArr[0] = 42, and OutArr[1] = 43.
Instead, we get:

# | Name:            ExpectedOutArr
# | Format:          Int32
# | Stride:          4
# | ArraySize:       2
# | Data:
# |   - [ 42 ]
# |   - [ 43 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Got:
# | ---
# | Name:            OutArr
# | Format:          Int32
# | Stride:          4
# | ArraySize:       2
# | Data:
# |   - [ 43 ]
# |   - [ 0 ]

It looks like Out isn't being assigned correctly, since OutArr[1] is 0, but we expect it to be at least 1, since Tid.x should be 1 in that case.
This was found using the DXC compiler, with the d3d12 target, here:
https://github.com/llvm/offload-test-suite/actions/runs/29453229428/job/87483262159?pr=1090#step:13:580