initially needs a super-charge in order to conduct a distributed test with servers that dynamically come and go
then, conceptually, needs a transformation to decentralized distribution (aka coordicide)
The scope of this issue is pre-coordicide, ie building out the centralized coordinator element of trellis as it is, to:
enhance understanding of it currently and within the developing context of 0KN
enable it to have uptime (hence immortal), not just start-run-end, but a type of manager such that other servers (that manage their own configuration) are able to submit/register themselves to the coordinator
The coordinator component:
The scope of this issue is pre-coordicide, ie building out the centralized coordinator element of trellis as it is, to: