Phasing is similar between 2nd/2nd and 1st/3rd networks. The symmetrical network tends to work well if the drivers are similar, two direct radiators of about the same depth, for example. Asymmetrical networks are often better when the adjacent drivers are very different, such as is the case when one is a direct radiator and the other a horn, or if one is considerably longer than the other. You have summing to consider, and you want to try to keep 1/2λ cancellation from happening through the crossover overlap region in the targeted listening area.The Alpha 8 is pretty smooth through the midrange, so it sounds nice in a system with a low-order crossover, or one with no crossover at all. If it had more breakup, you would probably like it better with a higher slope. The tweeter will have increased excursion if run with a lower crossover slope, and this will make it less robust. I think it sounds better with a higher slope anyway. So I think I'd probably lean towards the 1st/3rd solution. You might stagger the crossover points of the two to get summing right.