diff --git a/csharp/src/Google.Protobuf.Test/JsonParserTest.cs b/csharp/src/Google.Protobuf.Test/JsonParserTest.cs
index 87a389aecbb5..69c9eb6e998b 100644
--- a/csharp/src/Google.Protobuf.Test/JsonParserTest.cs
+++ b/csharp/src/Google.Protobuf.Test/JsonParserTest.cs
@@ -551,6 +551,8 @@ public void NumberToDouble_Valid(string jsonValue, double expectedParsedValue)
         }
 
         [Test]
+        // Skip these test cases in .NET 5 because floating point parsing supports bigger values.
+        // These big values won't throw an error in the test.
 #if !NET5_0
         [TestCase("1.7977e308")]
         [TestCase("-1.7977e308")]
diff --git a/csharp/src/Google.Protobuf.Test/JsonTokenizerTest.cs b/csharp/src/Google.Protobuf.Test/JsonTokenizerTest.cs
index 55ec02ea023f..0cbc0a4ff8ec 100644
--- a/csharp/src/Google.Protobuf.Test/JsonTokenizerTest.cs
+++ b/csharp/src/Google.Protobuf.Test/JsonTokenizerTest.cs
@@ -199,6 +199,8 @@ public void NumberValue(string json, double expectedValue)
         [TestCase("1e-")]
         [TestCase("--")]
         [TestCase("--1")]
+        // Skip these test cases in .NET 5 because floating point parsing supports bigger values.
+        // These big values won't throw an error in the test.
 #if !NET5_0
         [TestCase("-1.7977e308")]
         [TestCase("1.7977e308")]
diff --git a/csharp/src/Google.Protobuf/WritingPrimitives.cs b/csharp/src/Google.Protobuf/WritingPrimitives.cs
index d33a298d3e7f..8beefc54c574 100644
--- a/csharp/src/Google.Protobuf/WritingPrimitives.cs
+++ b/csharp/src/Google.Protobuf/WritingPrimitives.cs
@@ -213,6 +213,8 @@ public static void WriteString(ref Span<byte> buffer, ref WriterInternalState st
             }
         }
 
+        // Calling this method with non-ASCII content will break.
+        // Content must be verified to be all ASCII before using this method.
         private static void WriteAsciiStringToBuffer(Span<byte> buffer, ref WriterInternalState state, string value, int length)
         {
             ref char sourceChars = ref MemoryMarshal.GetReference(value.AsSpan());
@@ -283,6 +285,9 @@ private static void NarrowFourUtf16CharsToAsciiAndWriteToBuffer(ref byte outputB
             else
 #endif
             {
+                // Fallback to non-SIMD approach when SIMD is not available.
+                // This could happen either because the APIs are not available, or hardware doesn't support it.
+                // Processing 4 chars at a time in this fallback is still faster than casting one char at a time.
                 if (BitConverter.IsLittleEndian)
                 {
                     outputBuffer = (byte)value;