create table tablename (runid int,rundate date,errorid int);

insert into tablename values 

(101   , '2017-04-11',  1),

(101   , '2017-04-11',  2),

(101   , '2017-04-11',  3),

(102   , '2017-04-22',  2),

(102   , '2017-04-22',  3),

(103   , '2017-04-26',  1),

(104   , '2017-04-27',  3),

(105   , '2017-04-28',  4)

;

​

with t1 as (select runId,runDate,errorID

    ,first_value(runDate) over(partition by errorID order by runDate desc rows between unbounded preceding and unbounded following) as last_seen

    from tablename)

select 

distinct t1.errorid

,first_value(t.rundate) over(partition by t1.errorID order by t1.runDate desc rows between unbounded preceding and unbounded following) as rundate

,first_value(t.runID) over(partition by t1.errorID order by t1.runDate desc rows between unbounded preceding and unbounded following) as runid

from t1

join tablename t on t.runDate>t1.last_seen