As I said, I do think being too strict on it would cause too much queue time. Honestly, I cannot possibly estimate what would lead to acceptable queue times, because I have no data on how many people are looking for a draft at a given time.
Much like normal matchmaking, the pool of players considered would grow larger as the queue time increases, but it should probably start larger than exacts ranks. For instance, it could put bronze + silver together, gold + platinum and diamond + mythic. Or, since I believe the gold population is significantly larger than the rest (I could be wrong), it could be bronze + silver, gold, platinum + diamond + mythic. Or it could be any other split that makes sense to someone who actually has the data.
Again, I cannot possibly know whether or not it's feasible from a queue time perspective, but I thought it would be an interesting discussion w.r.t. whether or not they should consider it in the first place.
"I cannot possibly estimate what would lead to acceptable queue times, because I have no data on how many people are looking for a draft at a given time."
Until a few days ago, we didn't have any of that data for player drafts either :)
This is an interesting question, with both queue time & gameplay implications. Queue time questions we can sort out as we get more data, but the gameplay questions are stickier. In short: one of the key virtues of Draft is the variability from pod to pod. The more similarly each pod you're in drafts, the more repetitive the format will feel.
This is something we talk and think about regularly. For now we feel like non-matchmade is the overall better experience, but we'll continue to monitor and discuss here.