I don't fully understand where the problem is. Where the bonds are supposed to be, S-Se bonds? Like this, those distances don't seem much longer than Cd-S.
In any case, you should not be troubled by the "sticks" connecting atoms in the visual representation: there are thresholds telling the program when to show them, and when not to show them. Actually, in GaussView, you can add sticks between the atoms you want, even if they are at 10 Ang distance.
Additionally, if you are using a Quantum Mechanical method (even if semi-empirical) sticks have no meaning at all, as they are only a representation for you.
What you should check is:
- the actual Se-S distance, and how it compares to literature values.
- whether the optimisation is proceeding correctly, and it is not oscillating between the same structures.
- possibly increase the level of theory if 1 is not satisfactory.
- if 3, use a suitable basis set for the metal, as probably "regular" basis sets are going to describe it poorly.