SQL Server 2019

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
create table tb(date date, phase char(1),user_id int);

insert into tb values
('1.1.20',	'a',	10),
('2.1.20',	'a',	10),
('3.1.20',	'b',	10),
('4.1.20',	'a',	10),
('5.1.20',	'a',	10),
('6.1.20',	'b',	10);

6 rows affected

with cte1 as(
        select user_id,date,
         case
           when phase = 'a' and (lag(phase)over(order by date) is null or lag(phase)over(order by date) = 'b') 
            then 1
            else 0
          end grp
        from tb),
cte2 as(
    select user_id,date, sum(grp)over(order by date) as rnk
    from cte1)
select user_id,min(date)start_date,max(date)end_date
from cte2
group by user_id,rnk


user_id
start_date
end_date


10
2020-01-01
2020-03-01

10
2020-04-01
2020-06-01


with cte1 as(
   select *,
      case 
        when phase = 'b' then 'end'
        when phase = 'a' and 
           (select top 1 phase from tb t2 where t2.date < t1.date order by date desc) = 'a' 
          then 'no'
        else 'start' end grp
    from tb t1),
cte2 as(
    select *,
       (select top 1 date from cte1 t2 where t2.phase = 'a' and t2.date = t1.date)start_date,
       (select top 1 date from cte1 t2 where t2.phase = 'b' and t2.date = t1.date)end_date
    from cte1 t1
    where grp = 'start' or grp = 'end'),
cte3 as(
    select *
    from cte2
    where start_date is not null
),
cte4 as(
    select *
    from cte2
    where end_date is not null
)
select t1.user_id,t1.start_date,t2.end_date
from cte3 t1

cross apply 

  (select top 1 *
  from cte4 t2
  where t1.date < t2.date) t2


user_id
start_date
end_date


10
2020-01-01
2020-03-01

10
2020-04-01
2020-06-01