However, getting all the features included (assemblies, full GTOL support in drafting, harnesses, ECAD integration etc) while maintaining a vaguely usable UI seems to be currently beyond what FOSS projects can deliver. The CAD market is growing and highly competitive - and even then the professionally-developed software has trouble getting these things right. A couple of weeks ago I was involved with a plastic part redesign because the ECAD -> ProE interface lost a couple capacitors in the translation and prototype parts ended up interfering.
FEA and CAM are two other areas that require a significant amount of effort to get 'right'. ANSYS (FEA) has been developed since at least the 1980's and is still either 'easy to use' or 'fully featured'. CAM (CNC pathing, etc) is critical for prototyping components quickly and accurately.
Add to all that, the fact that none of the open interchange CAD formats (STEP, IGES, STL for 3d) contain all/any of the metadata (parametric model data, dimensioning and tolerances, for example) that proprietary formats do. For CAD software to be useful, it must be able to interchange data with other systems - every company/effort will need to exchange CAD data with another at some point, to communicate with a partner, vendor, consultant etc.
Long story short, proprietary CAD packages are the best ones available, which should be the real concern for a manned space program (since people's lives are clearly immediately at stake). They have the experienced people in the jobs market and the extensive knowledgebase needed to work through day-to-day problems. I can just google something to find out how I can do what I need to in ProE, or walk down the hall and talk to the guy who's been using it for 20 years. If I needed to submit a bug request and wait for a response and work-around, I wouldn't be able to do my job.