Despite a post-delete event handler, after a bulk delete a Lucene full text index is out of synch with the database so that a Lucene search ends with an ObjectNotFoundException. Is there a way to prevent this?

yes and no.
The next version (in SVN) will ignore not found objects (actually LIE), so accessing an unavailable object will not longer be done.
But fundamentally, there is no way for the index to detect which bulk of objects have been destroyed (through a HQL delete). Either you know which one and you update the index manually (usually does not make sense in a bulk operation), Or you reindex on a regular basis.