Cloud will definitely take over in the long run but its going to take another decade to get to the 80%+ mark and blue prints to successfully transition major systems like ERP's with many systems integrated with it in companyies away from monolithic application development.  As the ERP systems change the speed of cloud adoption will continue to increase. (even if its just for specific services within the application)

A big accelerator, in my opinion, are companies building poor performing on-prem clouds. This exact scenario: "... the wandering performance levels of their systems. For executives used to knowing what the predictable level of performance was, even if poor, having systems (appear to) run randomly slower or faster can be maddening." is common place in companies as internal infrastructure team build their own virtual environments without the correct experience or infrastructure to support it.  What results is IT spending more money on hardware than necessary and hiring more people which begins to make the cloud more appealing.

In the end what is absolutely maddening is when a system has unpredictable levels of performance and erodes trust in IT and the application vendor. 

From the executive level they don't think about it as a database or on-premise issue... they see it as the applications fault and want to replace it. (More $)
The Application vendor see's it as a client issue, IT infrastructure or because client is on a 4 year old version of the application and refuses to upgrade.

As a results the Vendors are 100% motivated to go to the cloud because they can control the environment, who is putting data into the system and how (another major reason for system issues), support goes down, and as the final motivated, reoccurring revenue goes up.  The only downside to vendors is they need to retool their development teams and architect their platforms which will take time for legacy applications and come at a significant expense.