Deleting an unpublished publication behaves unexpectedly

[William Grant]

Deleting a publication before it's published can have a rather surprising effect: any previous published publication is left untouched, and no longer has anything to supersede it, so the deletion basically just undoes the last upload.

See the last two Lucid publications on https://launchpad.net/ubuntu/+source/libipoddevice/+publishinghistory for an example of this mess. 0.5.3-3.2ubuntu2 was uploaded and then deleted before it was published, leaving 0.5.3-3.2ubuntu1 around. But the ubuntu2's binaries were left around, superseding ubuntu1's. So the ubuntu1 source was present with the ubuntu2 binaries.

[Robert Collins (lifeless)]

Sounds like deleting a publication needs to delete its binaries too?