The model as a whole can predict well while individual coefficients swing wildly between samples and carry large standard errors.
The diagnostic: variance inflation factors, or simply noticing that coefficients flip sign when you add or remove a variable.
The trap. It does not bias predictions - if you only care about forecasting, it may not matter at all. It destroys interpretation of individual coefficients.
In finance it is everywhere, since candidate factors are usually correlated. Ridge regression handles it by shrinking coefficients, trading a little bias for much less variance.