Two points from me.
First, it is unlikely you'll safely make the runway in the case of power loss mid-downwind, no matter what your spacing. You might physically reach the runway you've just departed from, but you're pretty sure of running off the end (unless you happen to be training at Edwards AFB). Much more achievable is a turnback from ~500+ feet AGL after takeoff, or through crosswind, or early downwind. From mid-downwind onwards, on a regular training circuit you just won't make the runway. (There's a narrow window just before base turn, but that's discounted in this discussion as it's a highly unlikely scenario).
So we can discount trying to make the runway as a factor in circuit design. This brings me to my...
Second point. The primary reason for flying circuits is to give a student pilot the opportunity to practice takeoff, approach, and landing. Therefore, the circuit should maximize the student's training value, by being as efficient as possible while coonsidering the aircraft's performance characteristics. As student ability changes, so should the manner in which the circuit is taught/flown. In my opinion, the oval circuit is better, as it challenges judgment and assessment in ways the square pattern cannot, and it copes better with changing wind conditions.