Hi again,

So, summing up what we have so far (patch attached), the effect of these
changes is to make the errors go away, but i'm not sure if the behaviour
is expected: with the patch applied, A T will "work" for me in an
nnselect group, with the caveat that, when gnus-refer-thread-use-search
is t (or when calling gnus-refer-thread with C-u otherwise), the result
is limited to the email already present in the nnselect group... that
might be expected, in which case i think the patch can be applied (it's
basically what you already propose, Eric, except it constructs the query
as an alist).

Cheers,
jao
-- 
There are two hard things in computer science: cache invalidation,
naming things, and off-by-one errors.