Unstable simulation involving some X for some optimization flags

In this example, there is an unstable bug (unstable in the sense that most slight modifications of the source will correct it).

How to reproduce the bug:

bash run.sh

Examples of actions that correct the bug:

Moving the moveme assignment 4 lines up.
Assigning sig_a to 0 instead of in_data[89], which is anyways always 0.
Removing in_data[61:60] in the or-reduction whose value is assigned to sig_c.

Expected result:

Without -fno*
out_0: 0
With -fno*
out_0: 0

Actual result:

Without -fno*
out_0: 0
With -fno*
out_0: 1