One twin containing versions or multiple twins? What architectures are practical?

[Tom Berry]

Digital twins are used to help forecast or optimize systems in the future, whether short term or long term.
Conceptually this means a digital twin has past, present and possibly many future versions.

This raises questions on the computing architecture to support this. One way of helping to keeping these versions consistent is to have an architecture where "the" twin contains all the versions. I think it is more practical to have an architecture that allows multiple twins, each with its own set of versions. This would allow a simple architecture with different physical hosts for each twin: for example past=archive, present=as-built, future=study.

Any comments?
(This was one of the questions in todays ISO/IEC JTC1/SC41 Workshops on Digital Twins)