The 'repeatable' field type has a problem when used in a config form. Because the repeatable type has an extra 'fieldset' inside the the main 'field' tag and because JForm::getFieldsets does not exclude these specially nested 'fieldsets' and because this particular form's template doesn't use the joomla.edit.params layout (which includes a special exemption for fieldset tags with the repeatable attribute set to 'true'), when you place a 'repeatable' in a config form, you will get an extra tab that you didn't want. You can test this by inserting a 'repeatable' field in any config form. For example:

<field name="test" type="repeatable" label="Test" default="{'test':['']}">
    <fieldset name="test_modal" hidden="true" repeat="true">
        <field name="test" type="text" label="Test" />
    </fieldset>
</field>

paste this code in some appropriate location inside of any component's config.xml and then go to the administrator and look at the form. There should be an extra tab with nothing in it. This is the tab for the 'test_modal' fieldset. Apply this patch and it will go away.

This patch is kind of a quick fix for this issue. It solves the problem and shouldn't have any side-effects. In the longer term (because other things may be affected) more changes would be nice, including:

• The config form should be changed to make use of the joomla.edit.params layout.
• JForm::getFieldsets should (at least optionally) omit 'fieldset' tags nested within 'field' tags. This breaks BC but it was a kind of break of BC when 'fieldsets' began to be nested within 'field' tags in the first place because that was certainly not the expectation when JForm::getFieldsets was written.
• Possibly rethink the 'repeatable' field type. Although it serves a valuable purpose, almost everything about it is a mess. The xml to set it up is not so easy to remember, the layout of the fields on the rendered form is a bit limiting, the fact that you must 'save' the modal and then remember to 'save' the actual form can be confusing, using the stored values of the sub-fields is not as easy as it should be... Well, it's probably too late to do anything about any of that anyway.