Output tiff size mismatch for consecutive runs

Hi,

I’m running the SNAP 7 toolbox to process S1 GRD scenes. The output tiff size is (slightly) different for consecutive runs. Is this expected behavior? Is there anything to be worried about?

Input : S1A_IW_GRDH_1SDV_20210307T001201_20210307T001226_036885_04568E_C73D

Full workflow:
Apply orbit file > Remote GRD Border Noise > Thermal Noise Removal > Calibration > Terrain-Correction

Run 1:
1940319866 S1A_IW_GRDH_1SDV_20210307T001201_20210307T001226_036885_04568E_C73D_VV.tif

1962437547 S1A_IW_GRDH_1SDV_20210307T001201_20210307T001226_036885_04568E_C73D_VH.tif

Run 2:
1940319092 S1A_IW_GRDH_1SDV_20210307T001201_20210307T001226_036885_04568E_C73D_VV.tif

1962443002 S1A_IW_GRDH_1SDV_20210307T001201_20210307T001226_036885_04568E_C73D_VH.tif

(the output size is in bytes)

To clarify, the difference is not just in file size, but in actual data too.

What order of magnitude difference are we talking about here?

They are tiny, O(1e-7): https://code.earthengine.google.com/6bccd8d0189bd6a85326ba21f0fe0bc7

Looks like a bad double-to-float truncation somewhere. Still, this is a bit inconvenient, as it makes comparing before-and-after states harder.