Let's cover the first issue with Roll to Advance, predictability.
The only way I see to address this issue while retaining the rolling aspect is to use a bell curve distribution rather than a linear progression. Something along the line of roll 3D6 to advance and start the target at 18, give or take. That way, every roll has a chance to advance and the increase in chance to hit it next session is no longer linear. Sure, it is kind of a band aid, but it helps.
Eventually, the same problem creeps up at higher levels when the target goes significantly above 18, again requiring X sessions with no chance of advancement. It is acceptable because the chance to hit a level in one session pretty much disappears by mid level range. A low level party can luck out and hit the jackpot with a few thousand GPs and hit a level in one fell swoop, but doing that when you need 50,000 XP to level just won't happen.
By rolling 3D6, we can keep the advancement target at or near 18 for the first few levels to make them more dynamic.
My ideas for the next two issues (risk v reward and mixed parties) should also provide some more excitement to the whole process.
A revised chart of advancement targets is forthcoming.