diff --git a/arrow-array/src/builder/generic_bytes_view_builder.rs b/arrow-array/src/builder/generic_bytes_view_builder.rs
index deaf447d0e6e..770b5c8aae67 100644
--- a/arrow-array/src/builder/generic_bytes_view_builder.rs
+++ b/arrow-array/src/builder/generic_bytes_view_builder.rs
@@ -707,7 +707,7 @@ mod tests {
         exp_builder.flush_in_progress();
         assert_eq!(
             exp_builder.completed.last().unwrap().capacity(),
-            MAX_BLOCK_SIZE as usize
+            STARTING_BLOCK_SIZE as usize
         );
     }
 }
diff --git a/arrow-buffer/src/buffer/immutable.rs b/arrow-buffer/src/buffer/immutable.rs
index c53ef18ba58f..a8659ea64077 100644
--- a/arrow-buffer/src/buffer/immutable.rs
+++ b/arrow-buffer/src/buffer/immutable.rs
@@ -729,7 +729,7 @@ mod tests {
         let b = Buffer::from_vec(a);
         let back = b.into_vec::<i128>().unwrap();
         assert_eq!(back.len(), 0);
-        assert_eq!(back.capacity(), 20);
+        assert_eq!(back.capacity(), 0);
 
         // Test vec with values
         let mut a: Vec<i128> = Vec::with_capacity(3);
@@ -745,7 +745,7 @@ mod tests {
         let b = Buffer::from_vec(a);
         let back = b.into_vec::<i128>().unwrap();
         assert_eq!(back.len(), 7);
-        assert_eq!(back.capacity(), 20);
+        assert_eq!(back.capacity(), 8);
 
         // Test incorrect alignment
         let a: Vec<i128> = Vec::new();
@@ -779,7 +779,9 @@ mod tests {
         let mut b: Vec<i128> = Vec::with_capacity(5);
         b.extend_from_slice(&[1, 2, 3, 4]);
         let b = Buffer::from_vec(b);
-        b.into_vec::<i256>().unwrap_err();
+        assert_eq!(b.len(), 64);
+        assert_eq!(b.capacity(), 64);
+        b.into_vec::<i256>().unwrap();
 
         // Truncates length
         // This is an implementation quirk, but isn't harmful
@@ -808,7 +810,7 @@ mod tests {
         // Succeeds as no outstanding shared reference
         let back = slice.into_vec::<i128>().unwrap();
         assert_eq!(&back, &[1, 4, 7, 8]);
-        assert_eq!(back.capacity(), 20);
+        assert_eq!(back.capacity(), 8);
 
         // Slicing by non-multiple length truncates
         let mut a: Vec<i128> = Vec::with_capacity(8);
@@ -820,7 +822,7 @@ mod tests {
 
         let back = slice.into_vec::<i128>().unwrap();
         assert_eq!(&back, &[1, 4]);
-        assert_eq!(back.capacity(), 8);
+        assert_eq!(back.capacity(), 4);
 
         // Offset prevents conversion
         let a: Vec<u32> = vec![1, 3, 4, 6];
diff --git a/arrow-buffer/src/buffer/mutable.rs b/arrow-buffer/src/buffer/mutable.rs
index 7fcbd89dd262..5e6cae9fd0f5 100644
--- a/arrow-buffer/src/buffer/mutable.rs
+++ b/arrow-buffer/src/buffer/mutable.rs
@@ -341,7 +341,10 @@ impl MutableBuffer {
     }
 
     #[inline]
-    pub(super) fn into_buffer(self) -> Buffer {
+    pub(super) fn into_buffer(mut self) -> Buffer {
+        // Don't leak our extra capacity into the final buffer.
+        self.shrink_to_fit();
+
         let bytes = unsafe { Bytes::new(self.data, self.len, Deallocation::Standard(self.layout)) };
         std::mem::forget(self);
         Buffer::from_bytes(bytes)
diff --git a/arrow-flight/src/encode.rs b/arrow-flight/src/encode.rs
index 59fa8afd58d5..1804727edb76 100644
--- a/arrow-flight/src/encode.rs
+++ b/arrow-flight/src/encode.rs
@@ -1428,7 +1428,7 @@ mod tests {
         ])
         .unwrap();
 
-        verify_encoded_split(batch, 112).await;
+        verify_encoded_split(batch, 392).await;
     }
 
     #[tokio::test]