diff --git a/src/deb/mod.rs b/src/deb/mod.rs
index 1f7ae9b..1805d92 100644
--- a/src/deb/mod.rs
+++ b/src/deb/mod.rs
@@ -603,12 +603,34 @@ mod tests {
     /// NOTE: Ideally, we want to run this in CI, but it takes more than 1h
     /// to fully build the linux-riscv package on an amd64 builder, which is too
     /// much time
+    /// The series is the current LTS (26.04) rather than an interim one:
+    /// interim series vanish from the mirrors a few months after their EOL
+    /// (questing is already unreachable), an LTS stays pullable for years.
     #[ignore]
     #[tokio::test]
     #[test_log::test]
     #[cfg(target_arch = "x86_64")]
     async fn test_deb_linux_riscv_ubuntu_cross_end_to_end() {
-        test_build_end_to_end("linux-riscv", "questing", None, Some("riscv64"), true).await;
+        test_build_end_to_end("linux-riscv", "resolute", None, Some("riscv64"), true).await;
+    }
+
+    /// KNOWN-BROKEN cross build of the noble-era kernel, kept as an
+    /// ignored fixture to work from. Noble controls declare their build
+    /// tools unqualified (the `:native` idiom landed later), so exact dpkg
+    /// semantics demand host-architecture instances of them
+    /// (python3:riscv64, gcc-13:riscv64, clang-17:riscv64, ...) and the
+    /// resulting two-architecture install set is unsolvable: t64
+    /// libraries (libclang1-17t64) conflict with their own foreign-arch
+    /// variant, and the riscv64 toolchain instances drag depends chains
+    /// (gcc:riscv64) that do not resolve from the chroot sources. A real
+    /// run fails at the apt transaction with 'Unable to correct problems'.
+    /// Resolute-era controls declare :native properly; see the test above.
+    #[ignore]
+    #[tokio::test]
+    #[test_log::test]
+    #[cfg(target_arch = "x86_64")]
+    async fn test_deb_linux_riscv_noble_cross_end_to_end() {
+        test_build_end_to_end("linux-riscv", "noble", None, Some("riscv64"), true).await;
     }
 
     /// This is a specific test case for the latest gcc package on Debian