Oracle 21c

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
-- to save adding format to all the rows...
alter session set nls_timestamp_format = 'DD-MON-RR HH.MI.SS.FF AM'
alter session set nls_date_format = 'MM/DD/YYYY'
create table sampledata as
select 78328696 pkid, 12848815 customer_id, to_date('10/19/2022 11:05:38 AM','MM/DD/YYYY HH:MI:SS AM') actdate, 0.5 units, to_timestamp('19-OCT-22 11.05.38.947750000 AM') datetime from dual
union all
select 78328697, 12848815, to_date('10/19/2022 11:05:39 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.05.39.024819000 AM') from dual
union all
select 78328698, 12848815, to_date('10/19/2022 11:05:39 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.05.39.050859000 AM') from dual
union all
select 78321196, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.879037000 AM') from dual
union all
select 78321197, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.909837000 AM') from dual
union all
select 78321199, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.931040000 AM') from dual
union all
select 78321200, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.952084000 AM') from dual
union all
select 78321201, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.971703000 AM') from dual
union all
select 78321202, 12978419, to_date('10/19/2022 9:13:56 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.56.993092000 AM') from dual
union all
select 78321203, 12978419, to_date('10/19/2022 9:13:57 AM','MM/DD/YYYY HH:MI:SS AM'), 1, to_timestamp('19-OCT-22 09.13.57.014174000 AM') from dual
union all
select 78330838, 13710675, to_date('10/19/2022 11:44:29 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.44.29.465212000 AM') from dual
union all
select 78330839, 13710675, to_date('10/19/2022 11:44:29 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.44.29.498326000 AM') from dual
union all
select 78330840, 13710675, to_date('10/19/2022 11:44:29 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.44.29.527076000 AM') from dual
union all
select 78331625, 13710675, to_date('10/19/2022 11:56:28 AM','MM/DD/YYYY HH:MI:SS AM'), 0.5, to_timestamp('19-OCT-22 11.56.28.726815000 AM') from dual
14 rows affected

select max(pkid) as pkid,
  customer_id,
  trunc(max(actdate)) as accdate,
  sum(units) as units
from sampledata
match_recognize (
  partition by customer_id
  order by actdate
  measures match_number() as match_num
  all rows per match
  pattern (start_tran same_tran*)
  define same_tran as (actdate <= prev(actdate) + interval '2' second)
)
group by customer_id, match_num


PKID
CUSTOMER_ID
ACCDATE
UNITS


78328698
12848815
10/19/2022
1.5

78321203
12978419
10/19/2022
7

78330840
13710675
10/19/2022
1.5

78331625
13710675
10/19/2022
.5
PKID	CUSTOMER_ID	ACCDATE	UNITS
78328698	12848815	10/19/2022	1.5
78321203	12978419	10/19/2022	7
78330840	13710675	10/19/2022	1.5
78331625	13710675	10/19/2022	.5