The more I debug rigs with matrix connections, the more I long for the days of simple constraint connections. Are the nanoseconds gained by a matrix connection worth the aggravation of digging through a matrix network? To me, it's a resounding NO!!!
The old saying "can't teach an old dog new tricks" might apply to me. But I don't think so.
Constraint connections usually have easily understood, and identifiable hierarchical structures. Whereas matrix connections jump all over the place. Most times having to traverse multiple hierarchies to understand how things are driven. While it can be argued that it's a one-time point of aggravation. Experienced only while debugging. The gains will always be present when the rig is used.
I wonder if there is a workflow where constraints are used up to the point of releasing the rig into the wild. Convert constraints to matrix connections only at final approval of a rig. This way, for debugging, there's a simple constraint version of the rig that can be reconverted to matrix connections once fixes have been applied.