My most questionable design choice? Well, there's a few so let's break 'em down in honor of EN's impending completion.
One of the most oft-cited was the giant ambush spawn in Lyn's Tale. Basically, the entire enemy force got to move first, which drew some negative criticism from people more focused on design. Letting the enemies move first outright is, in my opinion, an interesting mechanic (moreso in a 30-chapter game with maps varying which faction moves first), but with ambush spawns already a point of contention, this might've taken things a bit too far. I've since made the spawn an ambush only on hard mode.
Karel's Tale is also kind of iffy from a sheer design perspective. It was originally inspired by playing the boss gauntlet in Super Smash Bros. Brawl, interestingly enough. One of those cases where something feels better conceptually than it does in practice. I added the first map in Bern as a response to criticism that the chapter felt like an extended cutscene, since "boss battles" are difficult to represent in Fire Emblem. There's only so much you can do to make a one-vs-one fight tough without either being too easy or too difficult; it veers from one to the other quite easily. Deranger suggested diversifying Karel's inventory, which adds a new layer of strategy to the gameplay, and I've since added onto that with additional item drops.
That same tale also features another mechanic that I enjoyed implementing in EN: the forced loss. This first appeared with Karel vs. Kaherdin, but that "forced" loss became the subject of an achievement later on. Similarly in Raven's Tale, I've since added a prompt to surrender due to criticism about the obscure nature of the chapter's ending condition. After surviving for the set number of turns, the map just got rid of the turn count. There's also an achievement incentive to survive ten more turns without surrendering, but the map left players with the most attractive option being: just let everyone die. It's a good narrative device, but a poor gameplay design in retrospect.
Karel's and Raven's tales are examples of the creativity enabled by EN's structural setup, but also show that just because you have a unique format that enables you to innovate with chapter design, that doesn't mean it'll always turn out well. There's always room to refine, and, in retrospect, I should've considered the implications of those designs more extensively from the outset (rather than designing around the narrative aspects and then addressing gameplay issues further down the road).