Each child thread should have consistent algorithm of data processing without any dependency on its pThreadID. The metadata modification during the processing is not possible due to locking that would eliminate parallelism.
The transformation between source dimension pDimPar and target equivalent dimension should be provided. If there are other dimensions similarly requiring transformation between source and target, the transformation should be provided with respect to the same rules as in case of pDimPar and its mapping - no thread should interact with target data that belong to another thread.