Oracle 23c

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.

with
 t(name,id1,id2) as (
    select 'A', to_date('01.01.2020','dd.mm.yyyy'),to_date('01.02.2023','dd.mm.yyyy') from dual union all
    select 'A', to_date('12.12.2021','dd.mm.yyyy'),to_date('20.05.2022','dd.mm.yyyy') from dual union all
    select 'A', to_date('12.12.2022','dd.mm.yyyy'),to_date('30.01.2023','dd.mm.yyyy') from dual union all
    select 'A', to_date('01.03.2023','dd.mm.yyyy'),to_date('01.04.2023','dd.mm.yyyy') from dual union all
    select 'A', to_date('01.01.2020','dd.mm.yyyy'),to_date('01.01.2024','dd.mm.yyyy') from dual union all
    select 'В', to_date('01.02.2020','dd.mm.yyyy'),to_date('01.04.2023','dd.mm.yyyy') from dual union all
    select 'В', to_date('01.05.2023','dd.mm.yyyy'),to_date('01.01.2024','dd.mm.yyyy') from dual union all
    select 'А', to_date('01.02.2024','dd.mm.yyyy'),to_date('01.03.2024','dd.mm.yyyy') from dual union all
    select 'А', to_date('01.04.2024','dd.mm.yyyy'),to_date('01.05.2024','dd.mm.yyyy') from dual
)
,time_points as (
    select distinct *
    from  t
    unpivot (dt for r in (id1,id2))
    order by name,dt,r
)
,intervals as (
    select *
    from (
       select name,dt as dt1,lead(dt)over(partition by name order by dt) as dt2
       from time_points
    ) i
    where exists(select * from t where t.id1<i.dt2 and t.id2>i.dt1 and i.name=t.name)
)
select *
from intervals i
match_recognize (
   partition by name
   order by dt1
   measures
      first(dt1) as start_
     ,last(dt2)  as end_
   pattern (a b*)
   define
      b as dt1=prev(dt2)
)

NAME	START_	END_
A	01-JAN-20	01-JAN-24
А	01-FEB-24	01-MAR-24
А	01-APR-24	01-MAY-24
В	01-FEB-20	01-APR-23
В	01-MAY-23	01-JAN-24