In that case the "unused" stack was overwritten by a function called on that thread. But I'd assume that Denuvo is careful to not call any third party code while it expects the "unused" stack data to remain unmodified, so this shouldn't happen here.
So this code can only break if the data is overwritten from code outside the control of that thread. On Unix, certain signals could cause that. Or the OS could decide to zero out the unused thread while the thread isn't running. Zeroing the thread could the helpful to wipe secrets spilled there (forward secrecy related), or if whole pages can be zeroed (via something like MADV_ZERO), it could reduce the memory consumption by allowing threads to shrink.
While I would like that thread zeroing feature, I think it's unlikely that MS will implement something like it. So the code should be unlikely to break in practice.
So this code can only break if the data is overwritten from code outside the control of that thread. On Unix, certain signals could cause that. Or the OS could decide to zero out the unused thread while the thread isn't running. Zeroing the thread could the helpful to wipe secrets spilled there (forward secrecy related), or if whole pages can be zeroed (via something like MADV_ZERO), it could reduce the memory consumption by allowing threads to shrink.
While I would like that thread zeroing feature, I think it's unlikely that MS will implement something like it. So the code should be unlikely to break in practice.