What is the flush step?

Its calling the flush function

It is truncating the database whenever it is initialized with that tie command and writes to the hash.

No , berkeleydb won't do that -- if truncation is occurring, its a problem with the code you did not show

So maybe I need to put a flush statement into an END block?

Sure, if an END block will get executed after you interrupt it, put it there -- see Signals and END block