We have issues such aswhich is waiting for a test to be written (and surely others that are still waiting to be discovered and diagnosed) while core remains broken. The problem in that specific case is a query in book.module...
$select = db_select('node', 'n')
... namely that the field parameter in the condition() call is unqualified. It should be
There are other similar cases, e.g. in blog.module, forum.module, even node.module, with unqualified column names like nid/uid/status/etc. in condition() calls, and I'm sure there are hundreds more in contrib.
We can try to fix these one by one, but they probably grow faster than we can fix them. I think it would make sense to require the field parameter to contain a qualified field name, with a table name or alias, at least for queries that are intended to be modified (
addTag('node_access')), but really for all queries.
Writing tests for node_access core issues is tedious, because AFAIK there's no node_access module in core — you'd have to write one for that specific test. Maybe it could be reused, but we need to test not only for ambiguous nid, but also for ambiguous uid, ambiguous status, etc.
The same goes for orderBy() and groupBy(), BTW. This needs to be fixed in DBTNG, not with tests.
(There's even bad documentation such as the hook_user_cancel sample code.)