One of my biggest hates in modern trends in UX was to have hidden controls, magic corners and symbols with overloaded meanings.
You had to have magical knowledge to use these controls. Windows 8 being the biggest offender with magic corners with no visual indicator that lead to essential controls needed for normal operation. Windows 8 made my blood boil the instant I tried to use it. Unity was near useless unless you knew the name of the application you wanted, and for those of us with dyslexia it was a near impossible interface to use.
This none-sense of form over function resulting in hiding function in obscure locations was always a doomed model. There was also the none-sense drive to unify touch interfaces and keyboard mouse interfaces as one UX experience. The fundamental issue here is the UX form grew out of the input devices we had.
punch cards -> card loader and status lights.
keyboard -> terminal interfaces
keyboard + mouse -> graphical windowed UI's
touch -> tiled interfaces with gesture controls.
The UX world was in love with the touch interface and believed it would be the only interface. "There can only be one." Basically every UX went this path. Windows, IOS, Windows mobile, Win 8.x/10, Gnome, Unity. OSX almost fell into this trap as well but at least Apple product tested this option and realised it sucked for certain device types.
When Ubuntu/Canonical decided they were going after the mobile market, tablet and phone they decided to completely ditch the ageing X system and rebuild the whole UX technology stake and model it around touch. In the Linux world the reaction was swift. Gnome3 with similiar ideas at around the same time. Over night MATE and Cinnamon were born. Linux Mint distro shot up the popularity charts to become arguably the most popular desktop distro. Gnome felt the hit the hardest with a mass defection of developer talent into the MATE and Cinnamon camps. This ultimately hurt Canonical as well developers simply avoided Unity.
Hopefully people will come to realise that the UX is tied to the input methodology/technology. Having two UX shells on a system is not the end of the world as a matter of fact I would prefer this model. Where the shell is tied to input device and or user preference.