This made me wonder as well, and so I did alot of digging in our company efficiency literature and the FCOM.
First, the FMGC dilligently budgets processor time, and doesn't recalculate Opt FL very often. Reenter the cruise level, make the box think, and you may very well see Opt FL change before your eyes.
Secondly the PROG page Opt FL is rounded to nearest 500 foot, so we can't "see" when our optimum level may indeed be above our present level, if only marginally. i.e., presently at FL360, displayed Opt FL 360, when in reality our "true" optimum might be FL362, with even slightly favorable winds up high the FMGC may want to step to FL380, confounding the well-intentioned aviator.
Thirdly, assuming other requirements are met, it's generally better to fly much higher than optimum than even slightly below it; i.e., presently at FL360, Opt FL 365, the FMGC will recomend a step to FL380. The box consistently prefers 1500 above Opt rather than 500 below, and performance data backs that up.
Lastly, we can't in practice fly the extremely slow climb that would be theoretically optimal. The vertical profile of our climb does indeed look like a stairstep, when transposed over the smoothly rising optimal. Since being below optimal is so prohibitive, the stair step doesn't average above and below the optimum slope, but instead biases to climb early.
Further, and you've no doubt already seen, how the step pays no heed to to the Rec Max FL (other than staying below it), in search of a strong tailwind up high and to a lesser degree ducking headwinds.