Since TRUNCATE has the same problems as REFRESH MATERIALIZED VIEW.

Use a table & view to fill it with data instead of materialized view. It
is similar in principle, but the table allows us to specify indices and
updates concurrently, instead of locking the whole table.

Wait until the DB starts accepting connections if it doesn't work at the
beginning. Blocking.

Detect when a database connection dies and reinitialize it.

As it is supposed to be faster by dropping it, it shall release some
resources in the server.

Recover from lost DB connection. Initialize a new one.

Disable the old connection without compression. It is no longer used, so
don't sit on the port.

Use sanity instead of assert. The advantage is it logs in a way that is
sent to the server. It may help find the crash-bug there somewhere.

Do the commit at the end. There's nothing gained by commiting during the
read.

As the flag is very slow during the updates.

If the database is very slow or blocked somehow, run only one instance.
There's no point in running several in a row when it unblocks. Also,
such thing stresses the DB unnecessarily.

If it is already running, don't start another instance. Or, more
exactly, terminate it right away.

Each item of the biflows table is updated once during its life, so leave
space for the updated entry there.

Fix bug causing the fake log to send after each event.

There seems to be some stack corruption. It shouldn't possibly in the
signal handler, but rewrite it all the same.

Call gdb on self to obtain backtrace and other info and put it to logs.

It timeouts a lot. See if increasing would help.

Include IP addresses of some TLD DNS servers often used for malware. Tag
them in the database.

Disconnect from DB for the time WHOIS queries are running. And try to
die in case of timeout, since the whois_query seems not to be doing
that.

Make sure the routers that have only firewall activity aren't skipped
over due to complexities in the query.

Don't think you're online unless you get a confirmation from the server.

Use sanity() instead of several cases of assert(), die() or abort().

Similar to assert, but with a message and not possible to exclude from
compilation. This is better for checking outputs, for example.

Since it signals broken message by returning NULL, not by calling die
directly.