The net protocol expects that the version-verack handshake
is complete before other messages are sent. There's no reason to
allow 100 out-of-order messages to come from a single peer before
deciding to ban them. This behavior should be treated more suspiciously
and ban the peer after one order of magnitude fewer messages.

To assess the impact of this patch, I examined mainnet logs spanning the last 4 months.
There were only 13 instances of missing-verack and none of missing-version.
All instances were received from unique peers. Given this data, I think it's safe to assume
we will not see 10+ missing-(version|verack) instances from any currently deployed nodes
that would otherwise not be banned prior to this patch.