The system performs a pre-processing step before running a simulation. In this step the trades are all valued once to determine the market variables that will be active in the simulation. The market variables are collected in a "rate cache".
The system then generates all of the scenarios, either from historic rates in the case of historic VaR, or using volatilities and
correlations for Monte Carlo credit. The rate cache now contains a 'vector' of rate shifts for each rate.
The vectorised valuations are performed for each trade.
The portfolio mark-to-markets are sorted and the scenario matching the confidence interval selected as the answer.
Nb: There is only one time step in a VaR calculation but there can be hundreds in a credit exposure calculation.