diff --git a/Makefile b/Makefile
index 65040e7a..98a00b0d 100644
--- a/Makefile
+++ b/Makefile
@@ -261,6 +261,7 @@ TESTS	= $(BIN)/test_smoke $(BIN)/test_lex $(BIN)/test_parse $(BIN)/test_check \
 	  $(BIN)/test_tagged_arr_variant \
 	  $(BIN)/test_nested_union_int_box \
 	  $(BIN)/test_tagged_assignable_cluster \
+	  $(BIN)/test_idcast_tagged_return \
 	  $(BIN)/test_inferred_array_global \
 	  $(BIN)/test_slice_str_global_arg \
 	  $(BIN)/test_slice_str_global_zero \
@@ -744,6 +745,12 @@ $(BIN)/test_tagged_assignable_cluster: test/wcc/836_tagged_assignable_cluster.c
 		$(LIB)/libwwrt.a | $(BIN)
 	$(CC) $(CFLAGS) -o $@ $<
 
+$(BIN)/test_idcast_tagged_return: test/wcc/837_idcast_tagged_return.c $(BIN)/ww \
+		$(BIN)/w6c $(BIN)/w6a $(BIN)/w6l \
+		$(BIN)/ww_ww $(BIN)/w6c_ww $(BIN)/w6a_ww $(BIN)/w6l_ww \
+		$(LIB)/libwwrt.a | $(BIN)
+	$(CC) $(CFLAGS) -o $@ $<
+
 $(BIN)/test_inferred_array_global: test/wcc/833_inferred_array_global.c $(BIN)/ww \
 		$(BIN)/w6c $(BIN)/w6a $(BIN)/w6l \
 		$(BIN)/ww_ww $(BIN)/w6c_ww $(BIN)/w6a_ww $(BIN)/w6l_ww \
diff --git a/selfhost/cmd/w6c/main.combined.ww b/selfhost/cmd/w6c/main.combined.ww
index f83275d4..3d1f86cd 100644
--- a/selfhost/cmd/w6c/main.combined.ww
+++ b/selfhost/cmd/w6c/main.combined.ww
@@ -35057,6 +35057,24 @@ fn cgreturn(c: *cgen, n: *node) void = {
 						if (ru1.kind == tykind.TY_TAGGED && ru1 != fu1) {
 							needswiden = true;
 						};
+						// S3/#35: a SAME-TYPE tagged CAST (ru1 == fu1,
+						// rhs an N_CAST) routes through the widener too.
+						// cstage keeps the N_CAST out of the srcreg
+						// passthrough (kind != N_CALL/INDEX/DOT) and the
+						// widen branch peels the identity cast internally
+						// (cg_widen_tagged_store cg_tagged_castpeel,
+						// cgen.c:2553), so cs emits the scratch-widen (NOT
+						// passthrough) for `return v: u` over an ident/
+						// call/dot source. ww's forwardtagged keys on
+						// rhs.kind (N_CAST uncovered), so the same-type
+						// cast fell to the scalar shuffle and synthesized
+						// tag 0 (silent). The N_CAST guard mirrors cstage's
+						// "N_CAST defeats srcreg"; peeling here instead
+						// would reroute a call/dot source to passthrough
+						// and break byte-id.
+						if (rhs.kind == nkind.N_CAST && ru1.kind == tykind.TY_TAGGED && ru1 == fu1) {
+							needswiden = true;
+						};
 					};
 				};
 			};
diff --git a/selfhost/cmd/wcc/cgenstmt.ww b/selfhost/cmd/wcc/cgenstmt.ww
index 2d9f7428..52d5d043 100644
--- a/selfhost/cmd/wcc/cgenstmt.ww
+++ b/selfhost/cmd/wcc/cgenstmt.ww
@@ -1059,6 +1059,24 @@ fn cgreturn(c: *cgen, n: *node) void = {
 						if (ru1.kind == tykind.TY_TAGGED && ru1 != fu1) {
 							needswiden = true;
 						};
+						// S3/#35: a SAME-TYPE tagged CAST (ru1 == fu1,
+						// rhs an N_CAST) routes through the widener too.
+						// cstage keeps the N_CAST out of the srcreg
+						// passthrough (kind != N_CALL/INDEX/DOT) and the
+						// widen branch peels the identity cast internally
+						// (cg_widen_tagged_store cg_tagged_castpeel,
+						// cgen.c:2553), so cs emits the scratch-widen (NOT
+						// passthrough) for `return v: u` over an ident/
+						// call/dot source. ww's forwardtagged keys on
+						// rhs.kind (N_CAST uncovered), so the same-type
+						// cast fell to the scalar shuffle and synthesized
+						// tag 0 (silent). The N_CAST guard mirrors cstage's
+						// "N_CAST defeats srcreg"; peeling here instead
+						// would reroute a call/dot source to passthrough
+						// and break byte-id.
+						if (rhs.kind == nkind.N_CAST && ru1.kind == tykind.TY_TAGGED && ru1 == fu1) {
+							needswiden = true;
+						};
 					};
 				};
 			};
diff --git a/selfhost/cmd/wwdump/main.combined.ww b/selfhost/cmd/wwdump/main.combined.ww
index d5d928b6..2a93d993 100644
--- a/selfhost/cmd/wwdump/main.combined.ww
+++ b/selfhost/cmd/wwdump/main.combined.ww
@@ -35057,6 +35057,24 @@ fn cgreturn(c: *cgen, n: *node) void = {
 						if (ru1.kind == tykind.TY_TAGGED && ru1 != fu1) {
 							needswiden = true;
 						};
+						// S3/#35: a SAME-TYPE tagged CAST (ru1 == fu1,
+						// rhs an N_CAST) routes through the widener too.
+						// cstage keeps the N_CAST out of the srcreg
+						// passthrough (kind != N_CALL/INDEX/DOT) and the
+						// widen branch peels the identity cast internally
+						// (cg_widen_tagged_store cg_tagged_castpeel,
+						// cgen.c:2553), so cs emits the scratch-widen (NOT
+						// passthrough) for `return v: u` over an ident/
+						// call/dot source. ww's forwardtagged keys on
+						// rhs.kind (N_CAST uncovered), so the same-type
+						// cast fell to the scalar shuffle and synthesized
+						// tag 0 (silent). The N_CAST guard mirrors cstage's
+						// "N_CAST defeats srcreg"; peeling here instead
+						// would reroute a call/dot source to passthrough
+						// and break byte-id.
+						if (rhs.kind == nkind.N_CAST && ru1.kind == tykind.TY_TAGGED && ru1 == fu1) {
+							needswiden = true;
+						};
 					};
 				};
 			};
diff --git a/test/wcc/837_idcast_tagged_return.c b/test/wcc/837_idcast_tagged_return.c
new file mode 100644
index 00000000..cdb806cd
--- /dev/null
+++ b/test/wcc/837_idcast_tagged_return.c
@@ -0,0 +1,190 @@
+/*
+ * 837_idcast_tagged_return — S3 / #35: a SAME-TYPE tagged CAST in a RETURN
+ * (`fn f() u = { ...; return v: u; }` where v: u — an identity cast).
+ * ken2 spec .ai/ken2-family-spec.md (S3 section). Category A, CORRECTNESS
+ * (widen-route), proven byte-id with cstage.
+ *
+ * ww's cgreturn keyed the tagged passthrough on rhs.kind; an N_CAST was
+ * uncovered, so the same-type cast fell to the scalar-shuffle fast-path and
+ * synthesised tag 0 — SILENT on any non-variant-0 value (a bool boxed as the
+ * int variant: ran the int arm -> 70, want 11). cstage routes the N_CAST
+ * through the scratch-widener (the N_CAST defeats its srcreg passthrough),
+ * which peels the identity cast internally (cg_widen_tagged_store /
+ * cg_tagged_castpeel). The fix adds the same-type N_CAST guard to cgreturn's
+ * needswiden block so ww takes the widen route too — NOT a top-level peel,
+ * which would reroute a call/dot source to passthrough and break byte-id.
+ *
+ *   row          | shape                                  | want
+ *   -------------+----------------------------------------+------
+ *   S3_idcast    | return x: u  (x: u, 2nd-variant bool)  | 11
+ *   S3_nocast    | return x     (same value, no cast)     | 11
+ *   S3_intcast   | return x: u  (x: u, variant-0 int=7)   | 7
+ *
+ * The 2nd-variant (bool) value is mandatory: a variant-0 (int) value boxes to
+ * tag 0 coincidentally right (the S1 trap). S3_nocast pins the no-cast route
+ * unchanged; S3_intcast pins the cast route on a variant-0 value.
+ *
+ * BYTE-ID: NEUTRAL. The newly-covered shape (same-type tagged CAST in return)
+ * is non-bootstrap — had the bootstrap contained it the byte-id gate would
+ * already be RED (pre-fix ww emitted tag-0 scalar-shuffle, cstage the
+ * scratch-widen). 990-997 untouched.
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/stat.h>
+#include <sys/wait.h>
+
+static int
+runwait(const char *cmd)
+{
+	int rc = system(cmd);
+	if (rc == -1) return -1;
+	if (WIFEXITED(rc)) return WEXITSTATUS(rc);
+	return -1;
+}
+
+struct row { const char *label; const char *src; int want; };
+
+static const struct row rows[] = {
+	/* S3 — same-type tagged CAST in return, 2nd-variant (bool) value.
+	 * Pre-fix ww ran the int arm (70); now the widen route boxes tag 1. */
+	{ "S3_idcast",
+	  "package main;\n"
+	  "type u = (int | bool | str);\n"
+	  "fn f() u = { let x: u = true; return x: u; };\n"
+	  "export fn main() i32 = {\n"
+	  "\tmatch (f()) {\n"
+	  "\tcase let n: int => return 70;\n"
+	  "\tcase let b: bool => return 11;\n"
+	  "\tcase let s: str => return 22;\n"
+	  "\t};\n"
+	  "\treturn 99;\n"
+	  "};\n",
+	  11 },
+
+	/* S3 — no-cast control: identical value, plain return. Pins the
+	 * forwardtagged/passthrough route unchanged. */
+	{ "S3_nocast",
+	  "package main;\n"
+	  "type u = (int | bool | str);\n"
+	  "fn f() u = { let x: u = true; return x; };\n"
+	  "export fn main() i32 = {\n"
+	  "\tmatch (f()) {\n"
+	  "\tcase let n: int => return 70;\n"
+	  "\tcase let b: bool => return 11;\n"
+	  "\tcase let s: str => return 22;\n"
+	  "\t};\n"
+	  "\treturn 99;\n"
+	  "};\n",
+	  11 },
+
+	/* S3 — cast route on a variant-0 (int) value: the widen route must
+	 * box tag 0 and carry the payload (7), not just coincidentally land
+	 * on tag 0. */
+	{ "S3_intcast",
+	  "package main;\n"
+	  "type u = (int | bool | str);\n"
+	  "fn f() u = { let x: u = 7; return x: u; };\n"
+	  "export fn main() i32 = {\n"
+	  "\tmatch (f()) {\n"
+	  "\tcase let n: int => return n: i32;\n"
+	  "\tcase let b: bool => return 11;\n"
+	  "\tcase let s: str => return 22;\n"
+	  "\t};\n"
+	  "\treturn 99;\n"
+	  "};\n",
+	  7 },
+};
+
+static int
+run_driver(const char *driver, const struct row *r, int i)
+{
+	char src[64], tmpdir[64], cmd[1024];
+	snprintf(src, sizeof src, "/tmp/idc_%d_%d.ww", getpid(), i);
+	snprintf(tmpdir, sizeof tmpdir, "/tmp/idc_%d_d_%d", getpid(), i);
+
+	FILE *f = fopen(src, "wb");
+	if (!f) return -1;
+	fputs(r->src, f);
+	fclose(f);
+
+	mkdir(tmpdir, 0755);
+	snprintf(cmd, sizeof cmd, "cd %s && %s build %s 2>/dev/null",
+	    tmpdir, driver, src);
+	if (runwait(cmd) != 0) {
+		fprintf(stderr, "row[%s]: build via %s failed\n",
+		    r->label, driver);
+		unlink(src); rmdir(tmpdir);
+		return -1;
+	}
+
+	const char *base = strrchr(src, '/');
+	base = base ? base + 1 : src;
+	char outbin[128];
+	snprintf(outbin, sizeof outbin, "%s/%s", tmpdir, base);
+	char *dot = strrchr(outbin, '.');
+	if (dot && strcmp(dot, ".ww") == 0) *dot = '\0';
+	int got = runwait(outbin);
+
+	unlink(src); unlink(outbin); rmdir(tmpdir);
+	return got;
+}
+
+int
+main(void)
+{
+	const char *bin = getenv("BIN");
+	if (!bin) bin = "out/bin";
+	char absbin[1024];
+	if (bin[0] != '/') {
+		char cwd[1024];
+		if (getcwd(cwd, sizeof cwd) == NULL) return 1;
+		snprintf(absbin, sizeof absbin, "%s/%s", cwd, bin);
+		bin = absbin;
+	}
+
+	char cdrv[1024];
+	snprintf(cdrv, sizeof cdrv, "%s/ww", bin);
+	char wdrv[1024];
+	snprintf(wdrv, sizeof wdrv, "%s/ww_ww", bin);
+
+	struct { const char *name; const char *path; int gated_on_existence; }
+	drivers[] = {
+		{ "cstage",  cdrv, 0 },
+		{ "wwstage", wdrv, 1 },
+		{ NULL, NULL, 0 },
+	};
+
+	int n = (int)(sizeof rows / sizeof rows[0]);
+	int total = 0, fail = 0;
+
+	for (int d = 0; drivers[d].name; d++) {
+		if (drivers[d].gated_on_existence
+		    && access(drivers[d].path, X_OK) != 0) {
+			fprintf(stderr, "idcast_tagged_return: skip %s (no %s)\n",
+			    drivers[d].name, drivers[d].path);
+			continue;
+		}
+		for (int i = 0; i < n; i++) {
+			int got = run_driver(drivers[d].path, &rows[i], i);
+			total++;
+			if (got != rows[i].want) {
+				fprintf(stderr,
+				    "idcast_tagged_return[%s][%s]: exit=%d want=%d\n",
+				    drivers[d].name, rows[i].label,
+				    got, rows[i].want);
+				fail++;
+			}
+		}
+	}
+
+	if (fail) {
+		fprintf(stderr,
+		    "idcast_tagged_return: %d/%d fixtures failed\n", fail, total);
+		return 1;
+	}
+	printf("idcast_tagged_return: %d/%d ok\n", total, total);
+	return 0;
+}