Cumulative DateDiff with dependencies

Question

I'm trying to calculate a cumulative datediff field, but with some dependencies. I want to have a cumulative datediff for each row in my table for each client, but when the DateDiff Previous ID > 31, i have to start at 0. My table is set up in the following way

I've tried a couple of things myself, but didn't give me the expected result. Lag function datediff previous row

WHEN [EndDate Previous] is null THEN [DateDiff Current]
WHEN [DateDiff Previous ID] < 32 THEN [DateDiff Current] + LAG([DateDiff Current]) OVER(PARTITION BY ClientID ORDER BY [Startdate]
ELSE [DateDiff Current]
END

The result, it only sums the current and previous DateDiff

Lag Function with DateDiff current + Lag(Running DateDiff)

WHEN [EndDate Previous] is null THEN [DateDiff Current]
WHEN [DateDiff Previous ID] < 32 THEN [DateDiff Current] + LAG([Running DateDiff]) OVER(PARTITION BY ClientID ORDER BY [Startdate]
ELSE [DateDiff Current]
END

Unfortunately it gives back null values

Last thing I tried was with a Sum function

WHEN [EndDate Previous] is null THEN [DateDiff Current]
WHEN [DateDiff Previous ID] < 32 THEN SUM([DateDiff Current]) OVER(PARTITION BY ClientID ORDER BY [Startdate]
ELSE [DateDiff Current]
END

The result I get is a totale SUM for this client

The final result I want shown below, but just can't get the right query for this, you guys got any ideas