We have recently released a survey on understanding the challenges that practitioners face with about software modeling and the particular problems that practitioners wish the researchers in academia to deal with.
The survey takes 2-7 minutes at most, and your contribution to the survey is extremely valuable and highly needed. Can you please support our research by filling in the survey and forwarding the survey to any of your stakeholders whom you think may wish to participate in the survey ?
The survey results will be made available online and published as a scientific journal/conference paper.
I took the survey to see the kind of questions you had in there. My closing comment is really the only relevant answer to how I work:
I honestly don't do modeling that much. I work directly in code and only use high-level diagrams on whiteboard as starting point straw men for discussion and exploration, then as a temporary guide. The main guide in development and design is code. See Jack Reeves' essays on "Code as Design" - it is very important for academia to understand and teach this perspective to the next generation of software developers.
Specifically, detailed design is guided and informed by unit test code. Higher level design is informed and guided by acceptance tests, load tests, integration tests, etc. in other words, tests are living documentation and executable specifications. The more a test framework allows you to express the design as executable code, the more it allows for refactoring, the more it supports my work and therefore the more I will use it. If they are just pretty drawings that I have to keep updating as the software design evolves, then models serve very little use for me and they will only be used after the fact as a way to document what was created rather than as a way to facilitate development.