You are correct about the replication problems, but this also varies heavily depending what scientific discipline you look at.

Also if you do science you may take the results oft another scientist (if they make sense and are peer revievewed) and build your next experiment on it, which may also work out and get peer reviewed.

So even with the replication problem science can work and build on thousands of experiments. But it would be better and needed that the experiments were reproducible.

You OK?

I am German, but I feel foremost European