Stumbled upon an excellent redundancy-focused blog post (HT: High Scalability). Here are just a few important points:

• Don’t make things too complex;
• Don’t add more risk than you take away;
• You’ve got to fail over in the right direction;
• You must be able to return to fully-redundant mode.

I’m guessing that people promoting stretched VLANs, vSphere and/or NSX clusters running across multiple sites, weird combination of EVPN and OTV, and a dozen similar shenanigans never considered any one of these points.