The page-partial attribute should not be necessary
Currently, the page-partial attribute must be added to any page that gets included in another page (whether entirely or via tags) in order to guarantee the source gets included and not the converted output. This attribute is necessary since, otherwise, Antora doesn't keep the source of the page around after converting the page. This was done to conserve memory. However, it may be an unnecessary or premature optimization, and it causes confusion for users.
First, test the performance with several large sites to see if keeping the AsciiDoc source around after conversion even affects the memory. If it does, we may have to reconsider this change, or figure out how to localize it.
The source should still be purged in the default pipeline after all pages have been converted.
The workaround is to set
page-partial in the playbook, which sets the attribute by default on all pages. If we determine that setting it by default is too costly, document this workaround.
Note that this change will not break any existing sites that already set page-partial per page (aka a la carte). That attribute will simply be ignored.