Just some background: Here I wrote “I definitely want to distribute the responsibilities, so that the project can better scale”. Allowing to change the build system was a first example of such a “distribution”. Although I personally always voted to stay with the existing Tycho build, I felt being almost alone with that opinion and thus accepted the decision of the other maintainers to go for the bnd approach, which was then started directly after 2.5.0.M1.
Well, fact is that by the build changes the release build broke and we are not even technically able to build a new milestone yet. I am trying to push getting back into this position. But even then, the current state is indeed not good enough that I would consider it ready for another milestone build.
Unfortunately, I have the same experience and I fully understand the frustration. I often see many errors on startup and different ones on every startup which doesn’t really make it easy to report dedicated issues as it is unclear what triggered the problems, how to reproduce them etc.
I thus beg all @maintainers to help working on such issue and get back to a stability we had with pre-2.5 snapshots. Every maintainer should use latest snapshots make sure it works at least for their personal setup. Everyone else can help by reporting issues with exceptions that they come across.
To be not too negative here: The code indeed went under a huge restructuring over the past months and the maintainers did a immense effort on this. It cannot be avoided that there are side effects from this and this will be a one-time situation only. Once the issues are resolved, I agree with @David_Graeff that the stability of the openHAB code base should be even better than before as different kinds of issues and regressions can be avoided by the new structure. But until then, it must be indeed a priority to fix the current issues.
Good news is: I personally updated my production (!) system yesterday to the latest snapshot and it is working more or less smoothly.