Running total to retain previous row total instead of resetting to zero when a specific condition is met

Your best bet will be to use a recursive Common Table Expression.

I voluntarily chose the simplest use (row-by-row iterations), because it will be a lot to learn at first,
but you will then be able to optimize it (for example by using a windowed sum() over () to make a first pass containing "everything surely under the cutoff" instead of the row-by-row iterations),
or use RDBMS specific functions (PostgreSQL User-Defined Aggregates).

Here is it with PostgreSQL (you didn't tell which RDBMS you use, so I choose the one I prefer…):

with recursive -- Give each row its position, we'll need it to iterate from row to row: p as (select row_number() over() pos, * from t), -- Now "recurse" (iterate) over rows: r as ( -- Init our recursion with the first amount, use its amount as the first value of the running sum, and as the capped sum if it's less than the cap: select *, amount rs, case when amount > 5000000 then 0 else amount end capped from p where pos = 1 -- Now the recursive part: union all select p.*, rs + p.amount, capped + case when capped + p.amount > 5000000 then 0 else p.amount end -- Use r itself (the previous iteration of our recursive CTE) to have the pos of the last iteration, and fetch the row having the next pos from r join p on p.pos = r.pos + 1 ) select * from r order by pos; 

I've put it running on your data in an example fiddle you can play with.