I believe the fallicy in your argument is that "most people" that purchase software do not do so with the understanding that they are looking for a something that does not work as expected. Instead, they encounter some oddity as they use the program and at some point determine that something isn't performing as it should - Vyger's recent find concerning lost geocoding following a GEDCOM transfer is but one example. It's not that they don't worry about it as much as they do not expect the program to fail. Again Vyger's response to finding this problem and how it affected his opinion of the software illustrates this position.
"Most of the bugs" meaning people can use it and generally not worry about anything. Anyway just look how long it takes for bugs to crop up. It has been released with many people using it and looking for bugs and just now Vyger found the bug. Often you have to work for a long time to find bugs.
Had you ended this sentance after "but a lot", or "There are plenty", I would have agreed.
And I am not implying that there are no bugs, there are plenty, but a lot, most of the bugs have been fixed.
I am trying only to be a voice of reason. I find RM 3.2.6 to be an excellent genealogical program, and I suspect that RM4 will (in time) be its superior. However, I find its present flaws (and especially those that may yet to be identified) to make it a questionable program from the standpoint of data integrity. Futhermore, I believe it is misleading to diminish these issues and suggest that "most of the problems" have been identified. While that might be true, it has yet to be proven and while my opponents may not believe it, I really hope it's true...