Strange and inconsistent sensor and valve behaviour with gases and liquids (simple reproducible examples provided)

I think it's normal.

Look at the arrangement of the packets of polluted water (for example). You have 1 clean, 1 polluted, 1 hole.

Each time a clean water packed is in front of the sensor, the valve become inactive, so the water is blocked.

Then, a polluted packet comes in front of the sensor and activate it, so the clean water goes through the valve. Same for the polluted water, as intended by the setup. And then ... a clean water anew, blocked at the input of the valve, which create this hole.

This happens because [pipe sensor]-[pipe shutoff] setup does not exactly act like pipe filters. They don't properly do their jobs when any end is blocked while filters always work properly but consume more power. 

Since the element in the pipe is coming alternately in your setup, the pipe sensor keep sending on and off signals. Please keep in mind that shutoff and sensor is 1-tile away in this set up, and sensor sends its signal 1 frame after the element passes the sensor. (Edit: I feel like this is wrong after some observation, It's more like signal sending is instant, but shutoff activation is 1 frame behind)

And when you break the pipe, if the target element(in this case PW) happens to be in sensor when pipe broke up, the shutoff opens at next frame and let water pass through which happens to be in entrance of the shutoff. In the next frame, water passes the sensor but shutoff is still open so PW passes the shutoff. In the next frame, PW passes the sensor but shutoff is closed due to signal from previous frame, so nothing passes. These 3 steps are continuously repeated and that's what we see in the pic. 

About the 'rare' case, which only first pair of element passes and shutoff is permanently closed, I tested myself with your settings and found out when I broke the pipe and pipe is flowing again, element inflow into T junction is redetermined at that time, so in some case certain element goes into pipe in a row, instead of alternately. In that case, if 2 or more non-target element goes into pipe sensor in a row, shutoff is permanently closed.

To sum, although I don't know whether the developers have a plan to change this system or not, there are some methods to avoid the problems that you met. You have to make sure that any end of pipe is not blocked. Following pics are my settings on Aquatuners with pipe sensor and shutoff to avoid get damaged, hope you can find some help to get your solution here.

Edited November 28, 2018 by GnyoR