I am highlighting the sql logic i would use in database environment that i am looking to replicate here. It does not accept that logic.
Page 1 / 1
Max Features to Read works like your LIMIT logic.
As a workaround you could Max read your LIMIT + OFFSET (so 75000 features) and then use a Sampler to take the last 25000 features?
But there is probably some SQL database wizardry possible to avoid having to read those first 50000 features.