New startup logic:
a) list all the merged files starting at first-streamable-block and walk through them
b) if there is a HOLE (missing block file, followed by another file...), EXIT with clear error
c) the last merged-file + 100 becomes your start block for next bundle
d) load the content of your forkdb based on previous merged-files (back up to the LIB or to the first-streamable-block)
main loop logic:
when polling the one-block-files, if you don't have one-block-files in the range of your start-bundle but you have some "higher than that" (hole in one-block-files), EXIT with a clear error