Does clCreateProgramWithBinary really load kernels onto FPGA with 13.0 SDK ?

The queue scheduling would be responsible for that. So it's not that CvP is triggered by you calling any enqueuing function, it's when the need to replace the FPGA exists that the FPGA becomes reconfigured using CvP. For example if data movements are still occuring you don't want the CvP to kick in and reprogram the FPGA so the scheduler determines when it is safe to do so. So if you are using breakpoints to see when CvP is getting called up it could be a bit misleading as to when operations are actually occuring.

In general I think of the command queue as an asynchronous scheduler that I have no control over. Commands like buffer movements and kernel invocations occur in order but I have no control over when each occurs in time.