When verbosity level is more than 1 the test tries to print the script_exec_args passed to script. It fails when the tuple has more than 1 elements and is used in formatting with %r. The two tests fail as below :

./python -m test test_cmd_line_script -vv

======================================================================
ERROR: test_package_error (test.test_cmd_line_script.CmdLineTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/root/cpython/Lib/test/test_cmd_line_script.py", line 368, in test_package_error
    self._check_import_error(["-m", "test_pkg"], msg, cwd=script_dir)
  File "/root/cpython/Lib/test/test_cmd_line_script.py", line 148, in _check_import_error
    print('Output from test script %r:' % script_exec_args)
TypeError: not all arguments converted during string formatting

======================================================================
ERROR: test_package_recursion (test.test_cmd_line_script.CmdLineTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/root/cpython/Lib/test/test_cmd_line_script.py", line 379, in test_package_recursion
    self._check_import_error(["-m", "test_pkg"], msg, cwd=script_dir)
  File "/root/cpython/Lib/test/test_cmd_line_script.py", line 148, in _check_import_error
    print('Output from test script %r:' % script_exec_args)
TypeError: not all arguments converted during string formatting

Ran 44 tests in 4.936s

FAILED (errors=2)
test test_cmd_line_script failed
test_cmd_line_script failed

== Tests result: FAILURE ==

1 test failed:
test_cmd_line_script

Total duration: 5.0 sec
Tests result: FAILURE

I verified with master on Win10 debug 32 build. The relevant lines, 147-151, are

        if verbose > 1:
            print('Output from test script %r:' % script_exec_args)
            print(repr(err))
            print('Expected output: %r' % expected_msg)
        self.assertIn(expected_msg.encode('utf-8'), err)

Normally, 'script_exec_args' is singular, the absolute file name to be executed. In the two failing cases, file 'name' is an import name, preceded by -m. Line 148 prints an informational header, when very verbose is requested, that only serves to identify the following lines. It is *not* tested. Therefore, we need not worry about the exact format, just avoiding a spurious error. The following is equivalent except when the value is a list (which should have been a tuple), when is prints without error:

            print(f'Output from test script {script_exec_args!r:}')

"Output from test script ['-m', 'test_pkg']:" seems clear enough for any human reader. I ran entire suite with -vv and did not find any other new, unexpected errors. PR to follow shortly.

New changeset 7e711ea by Terry Jan Reedy in branch 'master':
bpo-41731: Make test_cmd_line_script pass with -vv (GH-22206)
7e711ea

New changeset 77901dc by Miss Islington (bot) in branch '3.8':
bpo-41731: Make test_cmd_line_script pass with -vv (GH-22206)
77901dc

New changeset 34e3c75 by Miss Islington (bot) in branch '3.9':
bpo-41731: Make test_cmd_line_script pass with -vv (GH-22206)
34e3c75

Thanks Terry.