discovery(correlation): evaluate decay-weighted causal chains

## Related epic

#5981 — discovery only; this issue does not block epic completion.

## What to build

Determine whether entity-keyed, time-decayed causal chains produce a trustworthy user-facing escalation view worth adding to the correlation product.

## Acceptance criteria

- [ ] Name the user problem, target surface, baseline, primary outcome, minimum acceptable result, and opportunity cost.
- [ ] Define allowed causal edge types, provenance, confidence semantics, temporal replay, and weakest-link behavior.
- [ ] Validate that Redis adjacency storage is sufficient and specify retention, invalidation, and recovery behavior.
- [ ] Test representative sanction-to-disruption-to-market chains and explicit non-causal negatives.
- [ ] Define the user-visible states for incomplete, disputed, stale, and unsupported chains.
- [ ] Produce an explicit promote, defer, or reject decision and require fresh implementation review before coding.

## Blocked by

- #5987

Start by reading blocked issue #5987 and the related discovery epic #5981. Evaluate the stated causal-chain, Redis storage, replay, negative-case, and user-state criteria, then document a promote, defer, or reject decision with implementation review required before coding.

Written by the indexing model from the issue text.