Mark Iroh adapters as deferred pending fork of iroh-ts

Update peer dep from @rayhanadev/iroh to @alkdev/iroh. Document
platform strategy: Linux native + WASM fallback, Windows/macOS
deferred. Fix outdated single-import reference in iroh-transport.md
to split spoke/hub imports. Resolve binding stability and NAPI/Deno
R&D items as covered by the fork path.

docs/architecture/event-targets/iroh-hub.md

@@ -6,8 +6,8 @@ last_updated: 2026-05-08
 # Iroh Hub Event Target
 
 **Import**: `@alkdev/pubsub/event-target-iroh-hub`
-**Peer dep**: `@rayhanadev/iroh` (optional, NAPI-RS native addon)
-**Status**: Not yet implemented. Needs R&D on binding stability, NAPI under Deno.
+**Peer dep**: `@alkdev/iroh` (optional, NAPI-RS native addon — pending fork of iroh-ts)
+**Status**: Deferred. Pending fork of iroh-ts with Linux + WASM platform targets.
 
 P2P QUIC event target for the hub (server) side. The hub accepts incoming connections and bidirectional streams. Manages multiple connected spokes.
 
@@ -78,7 +78,7 @@ dispatchEvent(event) {
 
 ## R&D Needed
 
-1. **Binding stability** — same as spoke adapter. `@rayhanadev/iroh` needs testing.
+1. **Fork of iroh-ts** — same as spoke adapter. Pending fork as `@alkdev/iroh`.
 2. **Concurrent accept** — can `endpoint.accept()` handle multiple simultaneous connections?
 3. **Stream vs. Connection per spoke** — current design: one bidirectional stream per spoke on a single connection. Alternative: one connection per spoke. Need to benchmark which is better for the expected workload.
 4. **iroh-gossip** — for true broadcast to many spokes, `iroh-gossip` would be more efficient than per-spoke streams. Not yet available in TS. The current subscription-tracked fan-out design works for moderate fan-out; gossip would be an optimization for very large fan-out later.