Verbose packets can be received at any point in time while packets are being
forwarded in testpmd, but output collection in interactive shells doesn't
support this behavior (or any asynchronous behavior for that matter). Currently
the way that output collection works with interactive shells is after sending a
command the framework will consume all output since the previous command that
is in the buffer. In the case of verbose output in testpmd, this means that if
you send any command while verbose packet forwarding is turned on, you will
consume and discard all of the verbose information that was sent to stdout
prior to that command being sent.

There are a few ways to handle this, one way that was proposed was just to
ensure that testpmd methods always return their raw output so that it can be
parsed, another was to track if verbose output is being sent and, if it is,
collecting it from every command that is sent so that it can be read later [1].
The second method seems preferred, but if it is done that way there should be
some method of time-stamping when each blob of verbose output was collected so
that developers have the ability to examine verbose output specifically between
two command, rather than all of the output gathered thus far.

[1]
https://inbox.dpdk.org/dev/CAKXZ7egonYtTivf-SZB=ixv+Utgpww8Sku9Jy8xj58mgV4KNwQ@mail.gmail.com/