diff --git a/MobileGL/MG_State/GLState/ProgramState/ProgramLinkTask.h b/MobileGL/MG_State/GLState/ProgramState/ProgramLinkTask.h
index 61b75424..46bd54f6 100644
--- a/MobileGL/MG_State/GLState/ProgramState/ProgramLinkTask.h
+++ b/MobileGL/MG_State/GLState/ProgramState/ProgramLinkTask.h
@@ -120,10 +120,9 @@ namespace MobileGL::MG_State::GLState {
             // L1 shader-translation memo key for this program's SPIR-V (see
             // MG_Util/ShaderTranspiler/TranslationCache.h). Built HERE, at the tail of phase
             // A, and not by phase B - two reasons, both structural:
-            //   * the key covers the four link-time request maps and the merged opaque
-            //     bindings, and one of those (explicitOpaqueUniformBindings) lives in
-            //     `artifacts`, which phase B is forbidden to read because the GL-thread join
-            //     moves it out from under phase B;
+            //   * the key covers the three link-time request maps, which live in `in` - and
+            //     it has to be built before the link, because a hit is what makes the link
+            //     unnecessary;
             //   * built once, it serves both the lookup and the insert, so the program's
             //     sources are copied into the blob exactly once per link.
             // Invalid (null blob) when the cache is disabled, or when a stage arrived
@@ -168,10 +167,11 @@ namespace MobileGL::MG_State::GLState {
         // with any other, and an attached shader that failed to compile. Split out of
         // ConsumeShaders so they still run - in the same order, with the same diagnostics -
         // BEFORE the L1 memo is consulted, rather than behind a hit that would skip them.
-        // The two lexical side channels the relaxed parse cannot provide, merged across
-        // stages. Reads the compile snapshots only, so it runs before any parse - the merged
-        // opaque bindings are part of the L1 memo key. Sets artifacts.infoLog and leaves
-        // linkStatus false when two stages disagree on an explicit uniform location.
+        // Merges the per-stage explicit default-block uniform locations glslang recorded at
+        // compile time. Reads the compile snapshots only, so it runs before any parse - and
+        // before the L1 memo, so a hit can never paper over a program that must fail to link.
+        // Sets artifacts.infoLog and leaves linkStatus false when two stages disagree on an
+        // explicit uniform location.
         void MergeShaderSideChannels();
         Bool ValidateAttachedShaders();
         Bool ConsumeShaders(Vector<SharedPtr<glslang::TShader>>& outShaders);
diff --git a/MobileGL/MG_State/GLState/ProgramState/ShaderObject.cpp b/MobileGL/MG_State/GLState/ProgramState/ShaderObject.cpp
index 86059076..8158e0f7 100644
--- a/MobileGL/MG_State/GLState/ProgramState/ShaderObject.cpp
+++ b/MobileGL/MG_State/GLState/ProgramState/ShaderObject.cpp
@@ -155,7 +155,7 @@ namespace MobileGL::MG_State::GLState {
         //
         // The failure case is covered too: the info log stays queryable because nothing is
         // cleared. And if the stored TShader already fed a link, the no-op leaves
-        // preprocessedSource and both side-channel maps intact, which is precisely what
+        // preprocessedSource and the explicit-location snapshot intact, which is precisely what
         // ClaimParsedShader's on-demand re-parse needs - a real recompile would have handed
         // the next link a fresh parse, the no-op hands it a fresh re-parse of the identical
         // source instead. Same result, one parse either way.
diff --git a/MobileGL/MG_Test/Program/AsyncSpirvPhaseTest.cpp b/MobileGL/MG_Test/Program/AsyncSpirvPhaseTest.cpp
index bcd91279..d0df8541 100644
--- a/MobileGL/MG_Test/Program/AsyncSpirvPhaseTest.cpp
+++ b/MobileGL/MG_Test/Program/AsyncSpirvPhaseTest.cpp
@@ -877,7 +877,7 @@ TEST_F(AsyncSpirvPhaseTest, AsyncOffAndAsyncOnProduceIdenticalSpirvAndShadow) {
     // The ASYNC arm runs FIRST, deliberately. Both arms must compile the same source text for
     // their SPIR-V to be comparable, and the first arm to run is the one that pays for the
     // cold path: it misses the per-context ShaderPreprocessCache and therefore executes
-    // PreprocessShaderSource, the reserved-identifier scan and both lexical extractions. Run
+    // PreprocessShaderSource and the lexical rejection scans. Run
     // the sync arm first and the async arm becomes a cache hit that never runs any of that on
     // a worker - which is exactly the half this case exists to compare.
     const SingleWorkerScope oneWorker;
diff --git a/MobileGL/MG_Test/Program/ProgramTest.cpp b/MobileGL/MG_Test/Program/ProgramTest.cpp
index 0827ab6e..a268b1e6 100644
--- a/MobileGL/MG_Test/Program/ProgramTest.cpp
+++ b/MobileGL/MG_Test/Program/ProgramTest.cpp
@@ -3169,9 +3169,9 @@ void main() { fragColor = vec4(pow(uBase, 2.2), 1.0); }
 //   Layer 1 - the same shader object re-sourced with byte-identical text keeps its
 //             compiled state, and glCompileShader on it is a no-op.
 //   Layer 2 - two DIFFERENT shader objects holding byte-identical text share the
-//             source-only half of the pipeline (preprocess + lexical checks +
-//             side-channel extraction) through the context's ShaderPreprocessCache,
-//             while each still gets its own glslang parse.
+//             source-only half of the pipeline (preprocess + the lexical rejection
+//             checks) through the context's ShaderPreprocessCache, while each still
+//             gets its own glslang parse.
 // ---------------------------------------------------------------------------
 namespace {
     const char* kP0bVs = R"(#version 330 core