Comment 1 for bug 1490711

Would it be possible to know that a machine FETCHED a kernel but then FAILED to boot it properly? If so, could we not automagically:

 * try the default release kernel, and if needed
 * try the next kernel up, and if needed
 * keep going through the HWE kernels, then
 * record the first one that works?

In this way, we could automagically "just work" when commissioning, automagically remember the needed HWE kernel, but still let someone set the kernel to use as a property of the machine if they want to do it manually?