Learning for CCIE… EIGRP surprises

Hello

Today something that surprised me. Let’s say we have an R4 with three paths to a given destination 10.10.10.0/24:

1. through R1 FD 10000, RD 2000, current successor
2. through R2 FD 20000, RD 5000, current feasible successor
3. through R3 FD 15000, RD 12000 !!!shorter than path through F2, but doesn’t meet the feasibility condition

My understanding of EIGRP was that if a path through R1 goes down, the feasible successor will be made the successor. It is NOT the case. What happens is as follows:

1. R1 goes down
2. R4 thinks: my path through R3 is now the shortest: let’s see if it meets the feasibility condition. It does not, so it puts the route in active state and sends a query to R3.
3. R3 replies.
4. R1 thinks now i have two paths: the shorter path is now through R3.
5. R1 makes the path through R3 to be the successor. Path through R2 is still the feasible successor.

shocking innit?