Multiple cointegrating vectors

[questionz]

1. If I have multiple cointegrating vectors (say, 3), how do I interpret the results? Do I include all the three vectors in the VECM model? That would give me
three different alphas for each equation. How should I interpret the results?

2. Should I use the same number of lags decided using VAR and lag length criteria before doing the cointegration test for my VECM model?

3. What if my results show non-normality? Is there anyway I can estimate VECM which takes account of non-normality of the data?

4. The heteroskedasticity results with cross terms and without cross terms are different? Which one should I choose? Also, my data shows serial correlation. Is there anyway I correct for hetero and serial correlation?

Answers to any of the above qns will be greatly appreciated.