For the pair factor
R = Z_3[eta]/(eta^2 - 3 eta),
the CM indecomposables are
B0, R, B3.
The only projective/free one is R, so in the singularity category D_sg(R) it vanishes.
Therefore the singularity category has exactly two indecomposable objects:
B0, B3.
From the matrix factorizations / syzygies,
Omega(B0) = B3,
Omega(B3) = B0.
Hence in the triangulated singularity category
Sigma(B0) = B3,
Sigma(B3) = B0,
Sigma^2 = Id.
So D_sg(R) is literally a two-object 2-periodic category.
The non-split CM exact sequences
0 -> B3 -> R -> B0 -> 0,
0 -> B0 -> R -> B3 -> 0
descend to the singularity category as the two irreducible odd arrows between B0 and B3.
So the AR quiver is
B0 <--> B3
with the translation swapping the two vertices.
In K_0(D_sg(R)), the exact sequence 0 -> B3 -> R -> B0 -> 0 gives
[B3] + [B0] = [R] = 0,
since R is perfect/projective and vanishes in the singularity category. Thus
[B3] = -[B0].
But Sigma(B0)=B3 and Sigma acts by negation on K_0, so
[B3] = -[B0]
again, and therefore the group is generated by one class of order 2:
K_0(D_sg(R)) ≅ Z/2Z.
For
O_sing ≅ Z_3(V7) × R,
the extra V7 branch is smooth/projective, so it disappears in the singularity category as well.
Therefore the full singular Hecke node has the same singularity category as the pair factor:
D_sg(O_sing) ≃ D_sg(R).
This is the most compressed form of the object so far: after modding out the smooth directions, the Hecke node is exactly a two-object 2-periodic singularity with K_0 = Z/2.