Power it only to read

Why switch it off

With 3V3 on VCC the whole time, a wet board passes a small direct current through the water from the VCC loops to the SENSE loops. Direct current through water is electrolysis: over time it dissolves metal from the positive side. That is why bare soil probes, which pass milliamps, are known for wearing out.

This board passes far less. Every path runs through the 1 MΩ, so the current is at most VCC / 1 MΩ: 3.3 µA from 3.3 V. And it only flows while water is on the loops. How long the gold-finished loops last under that has not been measured, so this book does not put a number on it. The page this replaces said "weeks"; nothing supports that for this board.

What can be said: the current only flows while VCC is on. Wire VCC to a pin instead of 3V3, switch it on to read and off again, and the loops carry current for half a second in every ten. That is a twentieth of the time.

A pin can power it

A GPIO driven HIGH sits at your board's logic voltage, which is what VCC needs. It has to supply what VCC draws: after the first moment, at most 3.3 µA. The first moment is the largest. With the loops shorted outright and the 100 nF empty, VCC charges the 100 nF through the 150 Ω alone: up to 22 mA from 3.3 V, 33 mA from 5 V, dying away in about 15 µs. That is the reason for the 150 Ω. The Uno's pins are rated for 40 mA absolute maximum, and real water, tens of kilohms and up, makes it microamps.

Then wait half a second

Here the 100 nF from why the capacitor matters. It starts empty each time VCC comes on, and fills through the water and the 1 MΩ.

Pick what is on the board and press Play. SIGNAL climbs with a time constant of at most 0.1 s, the longest when the board is nearly dry. Read at 5 ms, as the old page advised with its "few milliseconds", and a fogged board shows about a fifth of its value. Five time constants, 500 ms, brings every case within 1 % of where it is going. That is the wait the next sketch uses.