I think OAs/AoOs are connected with initiative and the grid, so that those choices necessarily influence each other. If you use 3e/4e style cyclical initiative and a grid, your possible OAs will look very different than if you use no grid and roll initiative each turn after actions are declared.
Personally, I like cyclical initiative because it's faster. I even use a single roll for the other side, so that after the first round (+ surprise round) the initiative can be resolved by side. This makes combat simple and lets slower players think while the others' actions are resolved. There's also no need to use the Delay action in 3e rules.
Example: rogue rolls 22, fighter 13, cleric 8 and mage 7, the zombies roll 10. First rogue and fighter go. Then zombies go. After that it alternates between the PCs and zombies without need to track individual order.
When I'm playing gridless, I would prefer not to need OAs at all, because I do it when I want to resolve the combat quickly. Spellcasting is the only thing that in my opinion needs it for balance (assuming cyclical initiative), so if that could be solved another way I'd be happy. One round casting time for most spells would do it, as I've mentioned elsewhere.
With grids, OTOH, I want OAs that encourage tactical actions. Maybe you should never trigger an OA from an enemy you attack with a melee weapon? This would make the difference between fighting retreat and running clear. The 3e rule on withdrawing actually made it work the other way, which seemed weird. This way the Wizard who is sucked into melee could hit with his staff to disengage safely, or run away triggering an attack (but get to start casting immediately).