I often find myself writing the N+1st version of the same paper: we might have a short periodic internal report, a longer report for the funding agency, and we're now writing a conference paper. As N gets larger, this gets more and more confusing.
While the early iterations were good, they had different purposes, lengths, and level of details. For example, the internal "work in progress" version of a paper might have many preliminary results, but almost no introduction. The conference version needs to have a more detailed introduction, but fewer (more focused) results.
Currently, I find that writing paper version N+1 requires starting with a rough copy/paste from version N, printing all version until N, and then looking in all previous versions for any missing information.
What are some strategies for better handing these multiple iterative versions of these papers?
Note that in contrast with this question, we don't start with a complete result (something that could be considered the "superset" paper). Early iterations are mostly rough drafts, and sometimes middle versions do diverge because of different audience/requirements.