fix: add health registry, shutdown manager, and wire all modules

HealthRegistry: global singleton with HealthStatus (Healthy/Degraded/
Unhealthy). Modules register health check closures at construction.
is_healthy(), is_ready(), to_json() for aggregated status.

Shutdown: global CancellationToken via OnceLock. install_signal_handler()
listens for SIGTERM/SIGINT. Modules listen on token.cancelled().

Wiring:
- KafkaTransport, GrpcTransport: register with health, Arc<AtomicBool>
- CircuitBreaker: register with health, maps Open->Unhealthy
- HttpServer: register with health, use global shutdown token
- TieredSink drainer: listen on global shutdown token
- ConfigReloader: listen on global shutdown token

Logging + config cascade added to all new transports (file, pipe, HTTP,
Redis). 663 tests pass.

Author: catinspace-au

Cargo.toml

@@ -20,9 +20,11 @@ categories = ["development-tools"]
 exclude = [".claude/", ".github/", "ci/", "ai/", "docs/", "examples/", "benches/", "scripts/"]
 
 [features]
-default = ["config", "logger", "metrics", "runtime"]
+default = ["config", "logger", "metrics", "runtime", "shutdown", "health"]
 
 # Core features
+health = []
+shutdown = ["tokio", "tokio-util"]
 runtime = ["dirs"]
 config = ["figment", "dotenvy", "serde_yaml_ng", "serde_json", "toml", "dirs", "tracing"]
 logger = ["tracing", "tracing-subscriber", "owo-colors", "serde_json", "tracing-throttle"]
@@ -104,7 +106,7 @@ secrets-aws = ["secrets", "aws-config", "aws-sdk-secretsmanager"]
 secrets-all = ["secrets-vault", "secrets-aws"]
 
 # Full feature set
-full = ["config", "config-reload", "logger", "metrics", "metrics-dfe", "otel", "otel-metrics", "runtime", "http", "http-server", "spool", "tiered-sink", "resilience", "database", "cache", "transport-all", "transport-grpc-vector-compat", "secrets-all", "directory-config", "directory-config-git", "deployment", "version-check", "scaling", "memory", "cli", "io", "dlq", "dlq-kafka", "output-file", "expression"]
+full = ["config", "config-reload", "logger", "metrics", "metrics-dfe", "otel", "otel-metrics", "runtime", "shutdown", "health", "http", "http-server", "spool", "tiered-sink", "resilience", "database", "cache", "transport-all", "transport-grpc-vector-compat", "secrets-all", "directory-config", "directory-config-git", "deployment", "version-check", "scaling", "memory", "cli", "io", "dlq", "dlq-kafka", "output-file", "expression"]
 
 [dependencies]
 # Serialisation (always needed)
@@ -158,8 +160,9 @@ metrics-util = { version = ">=0.20.1, <0.21", optional = true }
 metrics-exporter-opentelemetry = { version = ">=0.2.1, <0.3", optional = true }
 sysinfo = { version = ">=0.38.0, <0.39", optional = true }
 
-# Async runtime (for metrics server, http-server)
+# Async runtime (for metrics server, http-server, shutdown)
 tokio = { version = ">=1.50.0, <2", features = ["rt-multi-thread", "net", "sync", "time", "macros", "signal", "fs", "io-std", "io-util"], optional = true }
+tokio-util = { version = ">=0.7.14, <0.8", optional = true }
 
 # HTTP client — pinned to reqwest 0.12 until vaultrs and opentelemetry-otlp
 # support 0.13. reqwest-middleware 0.4 and reqwest-retry 0.7 target 0.12.

src/config/reloader.rs

@@ -277,6 +277,9 @@ impl<T: Clone + Send + Sync + 'static> ConfigReloader<T> {
 
     /// Main reload loop — waits for any trigger, then attempts reload.
     async fn run_loop(self) {
+        #[cfg(feature = "shutdown")]
+        let shutdown_token = crate::shutdown::token();
+
         // File polling state
         let mut last_modified: Option<SystemTime> =
             self.config.config_path.as_ref().and_then(|p| file_mtime(p));
@@ -308,15 +311,46 @@ impl<T: Clone + Send + Sync + 'static> ConfigReloader<T> {
         };
 
         loop {
-            let trigger = self
-                .wait_for_trigger(
-                    &mut poll_timer,
-                    &mut periodic_timer,
-                    #[cfg(unix)]
-                    &mut sighup,
-                    &mut last_modified,
-                )
-                .await;
+            // Check for global shutdown before waiting for next trigger
+            #[cfg(feature = "shutdown")]
+            if shutdown_token.is_cancelled() {
+                info!("Config reloader stopping (shutdown)");
+                return;
+            }
+
+            let trigger_result = {
+                #[cfg(feature = "shutdown")]
+                {
+                    tokio::select! {
+                        trigger = self.wait_for_trigger(
+                            &mut poll_timer,
+                            &mut periodic_timer,
+                            #[cfg(unix)]
+                            &mut sighup,
+                            &mut last_modified,
+                        ) => Some(trigger),
+                        () = shutdown_token.cancelled() => None,
+                    }
+                }
+                #[cfg(not(feature = "shutdown"))]
+                {
+                    Some(
+                        self.wait_for_trigger(
+                            &mut poll_timer,
+                            &mut periodic_timer,
+                            #[cfg(unix)]
+                            &mut sighup,
+                            &mut last_modified,
+                        )
+                        .await,
+                    )
+                }
+            };
+
+            let Some(trigger) = trigger_result else {
+                info!("Config reloader stopping (shutdown)");
+                return;
+            };
 
             // Debounce check
             if last_reload.elapsed() < self.config.debounce {

src/health/mod.rs

@@ -0,0 +1,29 @@
+// Project:   hyperi-rustlib
+// File:      src/health/mod.rs
+// Purpose:   Unified health registry for service health state
+// Language:  Rust
+//
+// License:   FSL-1.1-ALv2
+// Copyright: (c) 2026 HYPERI PTY LIMITED
+
+//! Unified health registry for service readiness and liveness.
+//!
+//! Provides a global singleton [`HealthRegistry`] that modules register
+//! into at construction. The `/readyz` endpoint (or any health check)
+//! queries the registry to determine overall service health.
+//!
+//! # Usage
+//!
+//! ```rust
+//! use hyperi_rustlib::health::{HealthRegistry, HealthStatus};
+//!
+//! // Register a component health check at construction
+//! HealthRegistry::register("kafka_consumer", || HealthStatus::Healthy);
+//!
+//! // Query overall health
+//! assert!(HealthRegistry::is_ready());
+//! ```
+
+pub mod registry;
+
+pub use registry::{HealthRegistry, HealthStatus};