SQL Server 2019

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
select *
into t
from (values (101, 'X1'),
             (103, 'Y2'),
             (104, 'Z1'),
             (101, 'Y1'),
             (102, 'X1'),
             (101, 'X3'),
             (103, 'Z1'),
             (101, 'X2'),
             (102, 'Y3'),
             (105, 'Y3')
      ) v(memberid, policy)
10 rows affected

select *
from t 
order by 1, 2


memberid
policy


101
X1

101
X2

101
X3

101
Y1

102
X1

102
Y3

103
Y2

103
Z1

104
Z1

105
Y3


with edges as (
      select distinct t.memberid as m1, t2.memberid as m2
      from t join
           t t2
           on t2.policy = t.policy -- and t2.memberid <> t.memberid
     ),
     cte as (
      select m1, m2, convert(varchar(max), concat(',', m1, ',', m2, ',')) as members, 1 as lev
      from edges
      union all
      select cte.m1, e.m2, concat(members, e.m2, ','), lev + 1
      from cte join
           edges e
           on cte.m2 = e.m1
      where members not like concat('%,', e.m2, ',%') 
     )
select m1, dense_rank() over (order by min(m2)) as grouping
from cte
group by m1


m1
grouping


101
1

102
1

105
1

103
2

104
2