There’s a similar topic eviciting vaults that may also be of interest.
I am slightly in favor of periodic random death of nodes (not just elders).
But I’m also in favor of changing age increments from exponential to linear (reason - “virtually all age will be roughly within the range 7-20”).
I’m also in favor of reducing the significance of elders with respect to message routing. The change in workload between adult to elder is probably too extreme (both the type of work and amount of work). I reckon an age-weighted probability for deciding who is in the message delivery groups would be more appropriate, so as nodes age their messaging load gradually increases. Getting off topic though so I’ll leave it at that, lots of pros and cons to each approach.
Just throwing some ideas around to chew on, would not want to introduce changes that delay the next testnet!