Detecting Cleanflight PID tuning issues with Blackbox: not enough P

This is third part of Cleanflight PID tuning tutorial with Blackbox. Previously I've showed examples of:

This time it is time for something slightly different: not enough P gain. Usually this problem can be identified without any log analysis. Symptoms are quite visible: multirotor is sluggish during maneuvers, has a tendency to change attitude on its own, constant course corrections are required. In worse cases, it is unflyable. But how does it look like on Blackbox logs.

First of all, symptoms are not so clearly visible. There are no huge oscillations for example. Zoomed out log might event look good on a first glance. For example like this:

On a lower section we see that P is doing something, D is not jumping, I is changing slightly. Even gyro line is nice and thin. That means there is no high frequency noise. But, two things might look suspicious: I term is slightly too active (is it compensating for low P?) and gyro trace is active although there is not input.

Let's zoom in to one section.

Nothing very suspicious here. Only I term looks quite active. Usually at 100% zoom it is flat. So, let's take a look at another moment of this flight:

Yes, here it it. Right in a middle. RC command goes up, but gyro trace goes down (!) and only after few moments it correct itself and start to follow input. Next, input goes to zero, while gyro is still above the line. Ver, very bad. What pilot (me) wanted to do? Pilot wanted to go forward. What quadcopter did? First it started to rotate back, then corrected itself to rotate forward and continued to rotate even when pilot did not wanted it to rotate anymore. And in total, it almost did not rotated at all. First move negated the latter.

Similar thing is visible here:

I is active, gyro is very active while smooth, RC input no so much. In other words: I tries to compensate, quad is unpredictable in flight. It was flyable, but every turn was a struggle. Not nice, not nice at all...