diff --git a/Changelog.rst b/Changelog.rst
index 3ec1572b6a..0635b7242e 100644
--- a/Changelog.rst
+++ b/Changelog.rst
@@ -11,8 +11,8 @@ version NEXT
   (https://github.com/NCAS-CMS/cf-python/issues/715)
 * Improve `cf.Field.collapse` performance by lazily computing reduced
   axis coordinates (https://github.com/NCAS-CMS/cf-python/issues/741)
-* Reduce output CFA file size with by tailoring the HDF5 chunk sizes
-  to fit exactly the CFA instruction variables
+* Reduce output CFA netCDF file size by setting the HDF5 chunksizes of
+  CFA variables to be no larger than required
   (https://github.com/NCAS-CMS/cf-python/issues/739)
 * Fix misleading error message when it is not possible to create area
   weights requested from `cf.Field.collapse`
diff --git a/cf/read_write/netcdf/netcdfwrite.py b/cf/read_write/netcdf/netcdfwrite.py
index 746ca73229..8890fe4da1 100644
--- a/cf/read_write/netcdf/netcdfwrite.py
+++ b/cf/read_write/netcdf/netcdfwrite.py
@@ -914,6 +914,7 @@ def _cfa_aggregation_instructions(self, data, cfvar):
         aggregation_format = []
         for indices in data.chunk_indices():
             file_details = self._cfa_get_file_details(data[indices])
+
             if len(file_details) != 1:
                 if file_details:
                     raise ValueError(
@@ -972,6 +973,8 @@ def _cfa_aggregation_instructions(self, data, cfvar):
             ):
                 n = n_trailing - len(filenames)
                 if n:
+                    # This chunk has fewer fragment files than some
+                    # others, so some padding is required.
                     pad = ("",) * n
                     aggregation_file[i] = filenames + pad
                     aggregation_format[i] = formats + pad