Resolve Client.Go-004..010 code-review findings

Client.Go-004: ran gofmt on alarms_test.go and galaxy_test.go; the tree is
now gofmt-clean.

Client.Go-005/009/010: migrated Dial/DialGalaxy off the deprecated
grpc.DialContext/WithBlock to grpc.NewClient via a shared dial helper, with
a DialTimeout-bounded readiness probe to keep fail-fast semantics; shared
callContext deadline arithmetic; updated the stale Dial doc comment. Test
harnesses use passthrough:///bufnet for the NewClient default-scheme change.

Client.Go-006: added GatewayError.Code() and an IsTransient(err) helper so
callers can classify transient gRPC failures.

Client.Go-007: newCorrelationID no longer returns an empty id when
crypto/rand fails — it falls back to a non-empty time+counter id.

Client.Go-008: added coverage_test.go for transport-credential resolution,
callContext deadline arithmetic, and native value/array edge kinds.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

clients/go/mxgateway/client.go

@@ -19,6 +19,7 @@ import (
 
 	pb "gitea.dohertylan.com/dohertj2/mxaccessgw/clients/go/internal/generated"
 	"google.golang.org/grpc"
+	"google.golang.org/grpc/connectivity"
 	"google.golang.org/grpc/credentials"
 	"google.golang.org/grpc/credentials/insecure"
 	"google.golang.org/protobuf/types/known/durationpb"
@@ -36,22 +37,36 @@ type Client struct {
 	opts Options
 }
 
-// Dial opens a gRPC connection to the gateway and configures auth metadata,
-// transport security, and blocking dial cancellation from ctx.
+// Dial opens a gRPC connection to the gateway and configures auth metadata
+// and transport security.
+//
+// The connection is created lazily with grpc.NewClient: the channel is not
+// established until the first RPC (or the readiness probe below) needs it, so
+// a gateway that is briefly unavailable at Dial time no longer turns into a
+// hard error — the connection recovers when the gateway comes up. To preserve
+// fail-fast behavior, Dial then runs an explicit readiness probe bounded by
+// DialTimeout (default 10s, or ctx's deadline when sooner): it triggers the
+// initial connect and waits for the channel to reach Ready, returning a
+// *GatewayError if the gateway cannot be reached in that window. Cancelling
+// ctx aborts the probe.
 func Dial(ctx context.Context, opts Options) (*Client, error) {
+	conn, err := dial(ctx, opts)
+	if err != nil {
+		return nil, err
+	}
+
+	return NewClient(conn, opts), nil
+}
+
+// dial builds the shared gRPC connection used by both Client and GalaxyClient:
+// it resolves transport credentials, assembles dial options, creates a lazy
+// connection with grpc.NewClient, and runs the DialTimeout-bounded readiness
+// probe so callers still fail fast when the gateway is unreachable.
+func dial(ctx context.Context, opts Options) (*grpc.ClientConn, error) {
 	if opts.Endpoint == "" {
 		return nil, errors.New("mxgateway: endpoint is required")
 	}
 
-	dialCtx := ctx
-	cancel := func() {}
-	if opts.DialTimeout > 0 {
-		dialCtx, cancel = context.WithTimeout(ctx, opts.DialTimeout)
-	} else if _, ok := ctx.Deadline(); !ok {
-		dialCtx, cancel = context.WithTimeout(ctx, defaultDialTimeout)
-	}
-	defer cancel()
-
 	transportCredentials, err := resolveTransportCredentials(opts)
 	if err != nil {
 		return nil, err
@@ -61,16 +76,46 @@ func Dial(ctx context.Context, opts Options) (*Client, error) {
 		grpc.WithTransportCredentials(transportCredentials),
 		grpc.WithUnaryInterceptor(unaryAuthInterceptor(opts.APIKey)),
 		grpc.WithStreamInterceptor(streamAuthInterceptor(opts.APIKey)),
-		grpc.WithBlock(),
 	}
 	dialOptions = append(dialOptions, opts.DialOptions...)
 
-	conn, err := grpc.DialContext(dialCtx, opts.Endpoint, dialOptions...)
+	conn, err := grpc.NewClient(opts.Endpoint, dialOptions...)
 	if err != nil {
 		return nil, &GatewayError{Op: "dial", Err: err}
 	}
 
-	return NewClient(conn, opts), nil
+	if err := waitForReady(ctx, conn, opts.DialTimeout); err != nil {
+		_ = conn.Close()
+		return nil, &GatewayError{Op: "dial", Err: err}
+	}
+
+	return conn, nil
+}
+
+// waitForReady triggers the initial connect on conn and blocks until the
+// channel reaches connectivity.Ready, the timeout elapses, or ctx is
+// cancelled. The wait is bounded by dialTimeout when positive, otherwise by
+// ctx's existing deadline, otherwise by defaultDialTimeout.
+func waitForReady(ctx context.Context, conn *grpc.ClientConn, dialTimeout time.Duration) error {
+	probeCtx := ctx
+	cancel := func() {}
+	if dialTimeout > 0 {
+		probeCtx, cancel = context.WithTimeout(ctx, dialTimeout)
+	} else if _, ok := ctx.Deadline(); !ok {
+		probeCtx, cancel = context.WithTimeout(ctx, defaultDialTimeout)
+	}
+	defer cancel()
+
+	conn.Connect()
+	for {
+		state := conn.GetState()
+		if state == connectivity.Ready {
+			return nil
+		}
+		if !conn.WaitForStateChange(probeCtx, state) {
+			return probeCtx.Err()
+		}
+	}
 }
 
 // NewClient wraps an existing gRPC connection. The caller owns closing conn
@@ -188,7 +233,15 @@ func (c *Client) Close() error {
 }
 
 func (c *Client) callContext(ctx context.Context) (context.Context, context.CancelFunc) {
-	timeout := c.opts.CallTimeout
+	return callContext(ctx, c.opts.CallTimeout)
+}
+
+// callContext derives a per-RPC context from ctx, applying callTimeout: zero
+// uses defaultCallTimeout, a negative value disables the bound entirely, and a
+// caller-supplied deadline that is already sooner than the derived timeout is
+// kept as-is rather than being lengthened.
+func callContext(ctx context.Context, callTimeout time.Duration) (context.Context, context.CancelFunc) {
+	timeout := callTimeout
 	if timeout == 0 {
 		timeout = defaultCallTimeout
 	}