Problems with HTML when exporting and importing XLIFF

[thomas-louis-osp]

I run a selfhosted version of Weblate (5.8.4) and I have some HTML strings (some of them with entities like &nbsp ;). Translation is done externally so we deliver XLIFF to the agency. It now turned out that under some conditions the HTML is converted into entities (e.g. &lt ;p&gt ; for "< p >"). This only happens when the HTML string contains Entities like &nbsp ;. Any other HTML content is exported with HTML tags. My expectations would be that HTML is securely enclosed in a CDATA section. But the more severe problem is the re-import of the XLIFF files. All strings containing entities or HTML tags are not imported (translation not found). This is a nightmare for us, because we either have to write an importer which uses the API or we have to manually add the translations.
I would like to know if this is a well known problem and if there is a more secure way to export and import HTML content in XLIFF files (I used both variants: XLIFF 1.1 and XLIFF with gettext Extension - no difference)?

Note: I had to put a space before the closing ; of the entities otherwise the entity gets resolved in this topic here