Kafka Streaming Suppress Feature to get hold of transactions which are late beyond grace period

I am currently using kafka streaming DSL suppress feature for day window. we might have situation where some of the events might come very late,beyond grace period.

As per kafka streaming documentation such events will be discarded which does not fit into window.

Please help me out.

1) Is it possible to get hold of such discarded events in the same flow?

Apache flink does provide hold of such very late events and would like to know if such feature available in streaming.

2) How feasible to hold intermittent aggregated data in the memory with DSL- suppress for day window considering millions of events flow through system?

Any timeline kafka streaming community will provide rockDB support soon to avoid application crash due to out of memory.