This data is from the simplest, quickest and easiest install possible. You have to start somewhere. There's no point building a Cray to see how well your binary adder works. Refinement will follow - if it is required.
Even so, this unsophisticated setup has delivered some interesting data.
From what I can estimate, thermal inertia is about 30 seconds for the current setup. Differentiating hot water events a minute apart has not been difficult with the data collected so far. Even then, if there are a number of short events close together, then the amount of hot water involved is not significant, so distinguishing each of them is unnecessary. If there is a need to get so particular, then the security footage is timestamped as well and it would be trivial to match traffic and temperature ... albeit a bit time consuming.
7 minutes is way, way outside any time constant resolution required. The first 30 seconds after water has stopped flowing is clearly visible and unambiguous. If water starts flowing again within that 30 seconds, then the inertia will still result in a small, but visible dip unless it was only off for a few of seconds. In that case, it is highly likely it is the same person ... and in the case of the (rather unlikely) need for full clarification in such a situation, security camera footage is available.
It would be good to get the water flow rate, but it's not clear whether such information is necessary to achieve the desired goal. That goal, by the way, is not to convict, but to initiate a conversation.
Please keep in mind the purpose of this exercise is not to track every millilitre of water and every millijoule of energy so you can bill each person for the cost.