Why does my plant face still say thirsty even right after I water it?

Sounds like two separate things stacked on top of each other, and both are fixable without rewriting anything. First, the physical side. A haworthia in a small pot is usually in a fast-draining gritty mix, and if the soil has dried out completely, it can go a bit hydrophobic. When that happens, water finds the gap between the root ball and the pot wall and runs straight down to the drainage holes. The bottom of the pot actually gets the least soaked that way, because the water is moving around the soil, not through it. So your sensor down near the bottom might genuinely be sitting in a spot that stays dry for hours while the middle of the root ball is fine. The fix is to water slowly, in circles, maybe two or three passes, and give it a minute between passes so the soil can actually absorb instead of channeling it out the bottom.

Second, check the actual number before blaming the face. Most of these projects print the raw moisture value to the serial monitor, and the one you describe streams data to a web dashboard too. Water it, then watch the number for ten minutes. If the number climbs quickly but the face changes slowly, your sketch has a rolling average or a delay built in, and there is often a smoothing value or a wait timer near the top of the code you can just shrink. If the number barely moves, it is placement, not software.

On placement: middle of the pot is not a bad spot at all, it is actually the textbook one. You want the sensing pad roughly at root depth, a bit off-center so it is not pressed against the pot wall, with soil packed snugly around it. Air gaps around the sensor make readings drift and respond slowly. For a 4 inch pot, about halfway down and slightly toward the edge is where the roots actually live. One more quirk to know about: capacitive sensors are much more sensitive along the face of the PCB than the tip, so 'push it in as deep as it goes' does not necessarily mean better contact. Snug soil contact along the sensing strip matters more than depth.

That face lying to you is annoying, I get it. I dealt with this exact issue when my client built a similar system for a succulent garden. It turns out the type of clay you use matters way more than the depth. If you use anything below 40% organic material, the moisture diff simply won't travel as far laterally. On my side of the street, I've found that even a 10% bump in ambient humidity can make the digital reading jump around til it settles. Try changing the potting mix ratio first.

I tried this with a diferent sensor for my office succulents and ran into something wierd. It wasn't the soil drying out, but it felt more like the chip wasn't checking the moisture often enought. My ESP8266 project was set up to refresh the face data every hour, and if a watering happened five minutes before the check, it felt like the face was wrong until the next cycle. Maybe the default timing settings are too slow for these projects? I am also having some issues getting the serial output to stabilize when the network is busy, wich is confusing things.

I think sometimes it isn't the code timing or the soil itself. When I ran a project like this, the issue turned out to be the power delivery. The chip draws a fair bit of current when it tries to read the sensor, and if the voltage dips even a little, the reading gets noisy or inconsistent. It is weird but happens. I ended up adding a seperate 3.3V buck converter just for the sensor module to keep the voltage stable during the measurement. That cleared up the strange delay we were seeing.