There are 2 problems I think here, or only one, but there are two different and non consistent behaviors.
I'll call the "scaling value" the 4th diagonal value in the transform matrix.

1st : With a volume

• Visualize a volume in the 3D window
• Create a new transform, and put the volume in the Transformed column.
• If you change the scaling value, nothing changes in the visualization.
• If you change any of the 3 other diag values, it changes something.

--> Hardened transform consistent with visualized transform, but both doesn't take in account the scaling value (4th diag value). I think they should.

2nd : With a model

• Visualize a model in the 3D window
• Create a new transform, and put the model in the Transformed column.
• If you change the scaling value, it changes in the visualization, as it should : value "10" downscale the model by 10.
• Now put 10 for every diagonal values. The visualization shows the same thing as if it was an identity matrix, which appears normal.
• Now harden this transform, the model is upscaled by 10, so it means that the scaling value hasn't been taken in account.

--> Hardened transform non consistent with visualized transform :

• Visualized : takes in account the scaling value
• Hardened : doesn't take in account the scaling value