The following returns garbage:
CREATE TABLE `phased_info` (
`animalid` varchar(18) DEFAULT NULL,
`chipid` tinyint(4) DEFAULT NULL,
`sha1sum` varchar(40) DEFAULT NULL
) ENGINE=Columnstore DEFAULT CHARSET=utf8;

// Default CHARSET=latin1 has same behavior

insert into phased_info values ('HO840F124135526679', 1, '1');
insert into phased_info values ('HO840F124135526679', 1, '2');
insert into phased_info values ('HO840F124135526679', 1, '3');
insert into phased_info values ('HO840F124135526679', 2, '1');
insert into phased_info values ('HO840F124135526679', 2, '2');
insert into phased_info values ('HO840F124135526679', 2, '3');

select group_concat(animalid),count(animalid) as ct from phased_info group by sha1sum;
--------------------------+

--------------------------+

--------------------------+
3 rows in set (0.15 sec)

However:
select group_concat(cast(animalid as char)),count(animalid) as ct from phased_info group by sha1sum;
--------------------------------------------------------------------------------------------------+

--------------------------------------------------------------------------------------------------+

--------------------------------------------------------------------------------------------------+
3 rows in set (0.18 sec)

and
select group_concat(animalid),count(sha1sum) as ct from phased_info group by sha1sum;--------------------------------------------------------------------------------------------------+

--------------------------------------------------------------------------------------------------+

--------------------------------------------------------------------------------------------------+
3 rows in set (0.06 sec)

So casting the column before group_concat, OR not using the same column in another aggregate, seems to fix it.