MicroStream stores persistent data in a physical form, typically in native file-system files.
Yes, as many as you like. Each MicroStream instance represents one coherent entity graph of persistent data.
Yes. This is already done automatically. The minimum and maximum size of every partial file can be configured, although this is a very technical detail that should not be relevant in most cases.
At any given time, only one JVM process may directly access the files representing a unique set of data. Such a restriction is crucial for the correct execution of any application: changes to an application's persistent data have to be guarded by the rules of the application's business logic, i.e. the process that currently runs the application. Allowing another process to bypass these rules would eventually result in catastrophic consistency errors. The requirement to distribute an application over multiple processes must be solved by a clustering approach (e.g. by distributing logic AND persistent data over multiple processes or by having one process to serve as the data master for multiple worker processes).