When running current python code, i get a memory-leak.
The xmldoc result seems to be corrupt or not compatible with new xmldocs after transformation.
This is problematicaly, because there are no way to join multiple transformed xmldocs without copying the nodes or reparse xmldocs. The libxml2 and libxslt binding are perfect for multiple transformations and join operations, because libxslt accept an xmldoc for transformation and also returns one which makes processing really fast. But with this bug this is an real performance show-stopper. Same bug was observed in the Perl-Bindings which is a hint that the bug must be in the c-library. I also tried to transform the doc which an transform_ctxt and the libxml2.XML_PARSE_NODICT option, but without success.

import libxml2
import libxslt

# XSLT Transform
doc = libxml2.parseDoc("""
<result>
	<result>abc</result>
	<result>def</result>
</result>
""")
styledoc = libxml2.parseDoc("""
<xsl:stylesheet version='1.0'
  xmlns:xsl='http://www.w3.org/1999/XSL/Transform'>

  <xsl:template match='/'>
  	<xsl:copy-of select="*"/>
  </xsl:template>
</xsl:stylesheet>
""")
xsl_transform = libxslt.parseStylesheetDoc(styledoc)
xml_result_doc = xsl_transform.applyStylesheet(doc, {})    
xsl_transform.freeStylesheet()
doc.freeDoc()

# Unlink Node from XSLT-Transformed result
xml_result_root = xml_result_doc.getRootElement()
xml_result_root.unlinkNode()
xml_result_root.setTreeDoc(None)
xml_result_root.freeNode()
xml_result_doc.freeDoc()