Null records create vastly different results

[Jim Van Overschelde]

I am flummoxed why I am getting vastly different results for these two queries when the second query simply eliminates one null record.

select count(distinct ID2), syear
from db.table1 a
where a.id2 is not null
and a.id2 not in
(select b.id2
from db.table2 b
group by b.id2
)
group by a.syear;
-- This query produces ZERO records. Subquery has 399,980 ID2s

select count(distinct ID2), syear
from db.table1 a
where a.id2 is not null
and a.id2 not in
(select b.id2
from db.table2 b
where b.id2 is not null
group by b.id2
)
group by a.syear;
-- This query produces counts totally 169k. Subquery has 399,979 ID2s

Why does including the one null record in the subquery produce no records in the main query?