Why is Snowflake returning records where complex WHERE clause is FALSE?

I have this query in Snowflake (which runs on a generator so it doesn't need a table if you want to just run it):

SELECT
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) AS date_column,
    DATEADD(week, -98, CURRENT_DATE) as anchor_date,
    current_date,
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) BETWEEN DATEADD(week, -98, CURRENT_DATE) AND CURRENT_DATE as truthy
FROM
    TABLE(GENERATOR(ROWCOUNT => 5000))
WHERE
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) BETWEEN DATEADD(week, -98, CURRENT_DATE) AND CURRENT_DATE
order by date_column

Which returns these records (and more):

Clearly it knows that the WHERE clause is FALSE for the first five records, but it returns those records anyway. Why? I'm guessing the table-free / generator setup may have something to do with it, but I can't figure out why.

I know I could wrap this in a SELECT * FROM () where truthy; - solving the problem isn't so much the issue as understanding why this doesn't work. What's going on?

so first to prove the SEQ4 are stable in the select as implied:

SELECT
    SEQ4() * 7 as a,
    SEQ4() * 7 as b,
FROM TABLE(GENERATOR(ROWCOUNT => 5000))
WHERE A != b

Query produced no results

good

now to use the results of the select region in the where:

SELECT
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) AS date_column,
    DATEADD(week, -98, CURRENT_DATE) as anchor_date,
    current_date,
    date_column BETWEEN anchor_date AND CURRENT_DATE as truthy
FROM TABLE(GENERATOR(ROWCOUNT => 5000))
WHERE truthy
order by date_column

change back to using a seperate expression in the where clause (how you have it):

SELECT
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) AS date_column,
    DATEADD(week, -98, CURRENT_DATE) as anchor_date,
    current_date,
    date_column BETWEEN anchor_date AND CURRENT_DATE as truthy
FROM TABLE(GENERATOR(ROWCOUNT => 5000))
WHERE DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) BETWEEN DATEADD(week, -98, CURRENT_DATE) AND CURRENT_DATE
order by date_column

I am going to guess the independant code is the same as explicitly using a sub-select:

SELECT
    DATEADD(day, SEQ4() * 7, '2022-11-13'::DATE) AS date_column,
    DATEADD(week, -98, CURRENT_DATE) as anchor_date,
    current_date,
    date_column BETWEEN anchor_date AND CURRENT_DATE as truthy
FROM (
    select SEQ4() as sub_seq
    from TABLE(GENERATOR(ROWCOUNT => 5000))
) WHERE DATEADD(day, sub_seq * 7, '2022-11-13'::DATE) BETWEEN DATEADD(week, -98, CURRENT_DATE) AND CURRENT_DATE
order by date_column

which has unwanted results which can be seen like this:

SELECT
    SEQ4() as a,
    a > 1000 as assumptions
FROM (
    select SEQ4() as b
    from TABLE(GENERATOR(ROWCOUNT => 5000))
) 
WHERE b > 1000
order by 2;

Now if the where clause here was a != b there are no unwanted results, as the rows are only processed once.