Hmmm... Strange requirement indeed. If I'm understanding this correctly, a quirky update might do it.

Set up your table like this:

create table #abc

(insertid int, cityid int, parentcityid int, cityname varchar(50), parentname varchar(50)

,PRIMARY KEY CLUSTERED (insertid, cityid, parentcityid))

insert into #abc

select 1,10,11,'A','B'

union all select 1,11,12,'B','C'

union all select 1,12,13,'C','D'

union all select 2,10,11,'A','B'

union all select 2,11,13,'B','D'

union all select 2,12,11,'C','A'

Then try this, but test it thoroughly against deeper cases to be sure that it works. It seems to for this limited test data anyway.

DECLARE @NextParentID INT = 0, @CityID INT = 0

UPDATE a

SET @NextParentID = CASE WHEN parentcityid >= @NextParentID THEN parentcityid + 1

WHEN CityID >= @NextParentID THEN CityID + 1

ELSE @NextParentID END

,@CityID = CityID = CASE WHEN (

SELECT CityID

FROM #abc b

WHERE b.InsertID = a.InsertID - 1 AND b.cityid = a.cityid AND

b.parentcityid <> a.parentcityid) IS NULL

THEN CityID ELSE CASE WHEN @CityID >= @NextParentID THEN @CityID + 1 ELSE @NextParentID END END

FROM #abc a

OPTION (MAXDOP 1)

SELECT * FROM #abc

DROP TABLE #abc