I guess the program in the main view is always syntactically correct? But I can still have a syntax error in my condition, so on what dimensions is this a net improvement vs just... typing the code?
I am a big fan of flow-based visual programming for signal processing domains (like GNU Radio), music and generative art (like Max/MSP). In those domains where there's a lot of data flowing through confusing but pure transformations, and having a GUI and being able to effortless inspect intermediary products and creatively try different pipelines of patches can be much more convenient than text programming.
But if you're just putting a lot of extra steps around a typical looking procedural scripting language, what's the win?