diff --git a/compliance_checker/acdd.py b/compliance_checker/acdd.py
index d3946fa7..831a4424 100644
--- a/compliance_checker/acdd.py
+++ b/compliance_checker/acdd.py
@@ -136,20 +136,6 @@ def get_applicable_variables(self, ds):
             if varname and (varname not in self.applicable_variables):
                 self.applicable_variables.append(varname)
 
-            # TODO?
-            # by assigning:
-
-            # self._applicable_variables = self.applicable_variables
-
-            # this entire if statement won't have to be called again;
-            # not assigning will result in some of th cached results being duplicated,
-            # and thus some check Results being duplicated. However, once assigned
-            # the cache is present for the entire life of the ACDD1_* checker object;
-            # this has unintended side effects on our unit tests which do not always
-            # (never, I believe) flush the cache between creating/loading new testing
-            # datasets. Because of this, cached objects from the previous dataset are
-            # kept in the object, causing downstream tests to fail.
-
         return self.applicable_variables
 
     def check_var_long_name(self, ds):
@@ -539,16 +525,16 @@ def verify_convention_version(self, ds):
         Verify that the version in the Conventions field is correct
         """
         try:
-            for convention in ds.getncattr("Conventions").replace(' ', '').split(','):
+            for convention in getattr(ds, "Conventions", '').replace(' ', '').split(','):
                 if convention == 'ACDD-' + self._cc_spec_version:
                     return ratable_result((2, 2), None, []) # name=None so grouped with Globals
 
             # if no/wrong ACDD convention, return appropriate result
             # Result will have name "Global Attributes" to group with globals
-            m = ["\"Conventions\" does not contain 'ACDD-{}'".format(self._cc_spec_version)]
+            m = ["Conventions does not contain 'ACDD-{}'".format(self._cc_spec_version)]
             return ratable_result((1, 2), "Global Attributes", m)
         except AttributeError: # NetCDF attribute not found
-            m = ["\"Conventions\" attribute not present"]
+            m = ["No Conventions attribute present; must contain ACDD-{}".format(self._cc_spec_version)]
             # Result will have name "Global Attributes" to group with globals
             return ratable_result((0, 2), "Global Attributes", m)
 
diff --git a/compliance_checker/base.py b/compliance_checker/base.py
index 39b91fdd..31cb2d27 100644
--- a/compliance_checker/base.py
+++ b/compliance_checker/base.py
@@ -242,6 +242,7 @@ def attr_check(l, ds, priority, ret_val, gname=None):
     :param netCDF4 dataset ds       : dataset being checked
     :param int priority             : priority level of check
     :param list ret_val             : result to be returned
+    :param str or None gname        : group name assigned to a group of attribute Results
     """
 
     msgs = []
diff --git a/compliance_checker/cf/cf.py b/compliance_checker/cf/cf.py
index f8e0a8d1..6961bb94 100644
--- a/compliance_checker/cf/cf.py
+++ b/compliance_checker/cf/cf.py
@@ -1032,7 +1032,7 @@ def _check_valid_udunits(self, ds, variable_name):
         should_be_dimensionless = (variable.dtype.char == 'S' or
                                    std_name_units_dimensionless)
 
-        valid_udunits = TestCtx(BaseCheck.LOW, self.section_titles["3.1"])
+        valid_udunits = TestCtx(BaseCheck.HIGH, self.section_titles["3.1"])
         are_udunits = (units is not None and util.units_known(units))
         valid_udunits.assert_true(should_be_dimensionless or are_udunits,
                                   'units for {}, "{}" are not recognized by UDUNITS'.format(variable_name, units))
diff --git a/compliance_checker/suite.py b/compliance_checker/suite.py
index 57cdbe6b..1fb5f93c 100644
--- a/compliance_checker/suite.py
+++ b/compliance_checker/suite.py
@@ -556,7 +556,8 @@ def process_table(res, check):
                     # separating this and the previous level
                     if has_printed:
                         print("")
-                    reason_str = "\n".join('* {}'.format(r) for r in reasons)
+                    # join alphabetized reasons together
+                    reason_str = "\n".join('* {}'.format(r) for r in sorted(reasons, key=lambda x: x[0]))
                     proc_str = "{}\n{}".format(issue, reason_str)
                     print(proc_str)
                     proc_strs.append(proc_str)