A: It may not be bad. Researchers often run collinearity diagnostics when they have no reason to do so. The only reason to worry about collinearity is that it gives you wide standard errors, which in turn give you wide confidence intervals and low-powered hypothesis tests. Wide standard errors make it hard to tell which variables, if any, are having an effect. If your standard errors aren't that wide, then you don't have a collinearity problem, and there is no reason to run collinearity diagnostics. If your standard errors are wide, then diagnostics can help you figure out whether collinearity is the reason. (There are other possible reasons for wide standard errors--for example, a small sample size.)
In most analyses, you aren't concerned about the effect of every variable. You have a few variables that you're interested in, and a bunch of others that you use as statistical controls. Collinearity among the control variables is often not a concern. You don't need to know which of the control variables is having an effect; you only need some assurance that they are accounted for in the model.
The worst case is that collinearity gives you wide standard errors for the variables of primary interest: You are writing a paper about the effect on Y of X1 and X2, and all you can say is that either X1 or X2 has an effect, but you can't tell which one or to what degree. Unfortunately, there's not much you can do about this. Perhaps the most common strategy is to omit X1 or X2, but this amounts to ignoring relevant information. The uncertainty caused by the collinearity is real. If people have been arguing about whether X1 or X2 is important, your analysis cannot settle the question.
Fox, J. (1991). Regression diagnostics. Thousand Oaks, CA: Sage.
Neter, J., Kutner, M.H., Nachtsheim, C.J., & Wasserman, W. Applied linear regression models (3rd ed.). Chicago: Irwin.