@Maciej54 The only way this is going to take traction is to start spreading awareness. So here's what im going to do.

Youtube, TikTok, Facebook groups including Home Assistant, Discord, X etc:

"Kasa KE100 disaster- AVOID THIS PURCHASE"

Expect a couple million hits overall and the negative PR might just push TPLink to do something.

The implementation is a disaster, a CS uni student can implement better hysteresis in about a week. The API is broken. It's absolutely pathetic

  @Afzal786 true, true.... but they don't seem to care. I made quite some effort describibg all issues in detail, reaching out to support, that have said they reached out to TP-link central and made them aware of the issue blah blah.... 

Whats wierd to me it seems only very few people (compared with the sales numbers) are actually complaining about it. 

Its a pitty, as from hgardware perspective these are quite nice devices, I also like the fact that the T315 temperature sensor is e-ink based. 

But well, they are 4x chaper than e.g. "TADO" devices, so you get what you paid for. 

  @Wayne-TP I believe there might be a slight change in operation since last year (*two firmware versions were released), but in general, as many users have noticed, the problem can be summarized as following:

The KE100 valves show unstable temperature regulation behavior, primarily due to inadequate hysteresis and overly aggressive control logic.

Instead of maintaining a steady, partially open position to compensate for system thermal inertia, the valves tend to oscillate between closed and open states.

This prevents the system from reaching equilibrium and maintaining a stable room temperature.

Observed behavior (for a 23 °C setpoint):

• When the temperature drops slightly below the target (e.g. 22.5 °C), the valve begins to open — often more than necessary.

• Due to the thermal inertia of the water-based heating system and room volume, the actual temperature rise is delayed.

• As the temperature starts to increase (around 22.7 °C), the valve begins closing prematurely.

• At about 22.8 °C, it typically closes completely, instead of making a minor proportional adjustment and waiting for stabilization. (the radiators are still hot, as they need TIME to dissipate the heat)

• then the radiators cool down, room starts to cool down.... temperature drops to 22.5 and the cycle starts over.

This lack of proper hysteresis or predictive damping results in continuous oscillation rather than proportional modulation.

In comparison, conventional TRVs are capable of stabilizing and remaining partially open (typically 20–30%) throughout a day or night to maintain temperature with minimal fluctuation.
The KE100 valves, however, repeatedly cycle between 0% and higher opening levels, causing noticeable thermal discomfort and audible flow noise.

If you would test the valve in a hypothetical system/room that has no inertia, they would most likely work. 

Many people report annoying "jigsaw" temperature graph as the valve is unable to reach stable operation.

I'll send the support ticket again nd let you know the number. 

Additionally I have noticed that even though all valves have good range and do not report interference - the ones in my bedroom that cause most issues seems to have bigger problems reading the temperature from the designated sensor. The fallback all the time to the built in thermometer, and because they are behind the curtains they go absolutely nuts. This is just an extra issue on top the poor temperature controll, that makes things even worse. Since I'm stuck with this system for now, I have ordered another hub to have a dedicated hub in that room.... lets hope it will improve the situation a bit...

First, see the range info - all good, right...? :

Now lets see difference in readings:

Temperature sensor: 22.7

Valve 1 - 24.4 (that is nuts - no wonder the valve goes nuts)

Valve 2 - 22.5 (that might be a delayed reading fro, the sensor)

And the infamous jig-saw heating pattern - these vales open and close like crazy. I understand that working behind the curtain might be more difficult, but a standard mechsnical TRV works just fine.... well, maybe it has to be set a bit higher:

  @Wayne-TP  yes, the readings seems to be stuck for some time. I did try to restart the app, or pull down to refresh, or go into the specific TRV. 
Also, note how much the temperature is off - it's normal for other TRV is another room to have 0.1C difference as they might read at different intervals, but 1.0C+ indicated that it made a fail over to internal sensor. 

Unfortunately my plan to get another KH100 did not work.... as TP-Link does not sell these as separate devices! I have, by mistake, ordered H100, that is the same hardware, but works only with TAPO! ahhhh how annoying! I cant use it with Kasa at all!

Anyway, this is a secondary issue (interference or unable to read the temp from designated sensor). 

The original issue is still more important,  that is excessive corrections of temperature that happen too quickly (sometime you can hear the vale makes correction every few minutes! - that can't work, as it will always fight against radiator inertia). That of course also leads to closing the vales too early as the previous correction was just fine, but it does another one too soon. (no night mode does not help either)