One of the challenges of doing this sort of thing non-professionally is that there’s no way to gather a bunch of people together at once. Many of the voice contributions come from people from other states. This makes it very hard to have to two characters say the same thing at the same time. When people are together they can sync up their timing and inflection instinctively. When they don’t even have a chance to hear each other it rarely sounds right. The tempo is too out of whack, or the stress is too dissimilar, or something.
I was actually pretty lucky in the previous episode where I got all three “um…”s to sync up. It helped that there’s not many ways to say “um”, Hannah said “um” several times, and Hermione submitted two takes. But generally the way I avoid this is by simply saying the repeated phrase in narration. It didn’t sound quite right with ‘Wingardium Leviosa’ in today’s episode, so I doubled Lavender’s line, offset it a bit, and changed the pitch so it sounded more like two different people.
It would be possible to coordinate multiple takes to get well-synced simultaneous lines, but it would be a fair bit more work for everyone involved. I don’t want to ask too much from my volunteers, and I need spare time to pursue my other interests as well, so I haven’t bothered yet. The work-around isn’t perfect, but it works well.