1. Is there a way to model a process, such that process instances will not keep

From my experience, the answer to this question is no. The more processes the more memory needed. However you can minimise the effect by using sub-process and storing less data in the main process and archiving/deleting sub-process before the main process.

With the process instances the answer I believe is yes. We had a situation where a large number of processes were brought back from the archives and massively expanded our KDB's. They are now far larger than they were previously and despite re-archiving the processes there is no changes to our KDB's. From my understanding there is currently no way of compressing back the KDB's once expanded, they leave 'white space' in the file which keeps the file size. Normally I believe that the white space is used by new processes but worth remembering if you have a massive increase in the number of processes that your kdb’s will be expanded and maintain that size, until more processes are require.

From my experience, the answer to this question is no. The more processes the more memory needed. However you can minimise the effect by using sub-process and storing less data in the main process and archiving/deleting sub-process before the main process.

With the process instances the answer I believe is yes. We had a situation where a large number of processes were brought back from the archives and massively expanded our KDB's. They are now far larger than they were previously and despite re-archiving the processes there is no changes to our KDB's. From my understanding there is currently no way of compressing back the KDB's once expanded, they leave 'white space' in the file which keeps the file size. Normally I believe that the white space is used by new processes but worth remembering if you have a massive increase in the number of processes that your kdb’s will be expanded and maintain that size, until more processes are require.