@joluinfante 

More info. This is "now" the temperature in two of three decos:

Hot air rises, of course... I've been looking into some kind of fan solution that activates automatically, controlled by a sensor. But I don't want to reinvent the wheel. Any ideas for cooling the set-top boxes without resorting to central air conditioning?

  @joluinfante 

Hi. I’ve read your post, and after performing a deep log analysis on an identical system (Deco XE75 Pro, firmware 1.3.0), I have some findings that might save you a lot of troubleshooting: the problem isn't the heat, even if the 37°C weather made it seem like the primary suspect.

What you are experiencing is a critical RAM exhaustion triggered by a firmware "infinite loop" that affects almost the entire Deco line.

Here is the technical breakdown of what is actually happening to your network:

1. The "Smoking Gun": The 802.11k Steering Loop

The error you see in your logs (tmpsvr: Error: memfree less than 10000 KB) is the system's "death rattle." When free memory drops below 10 MB, the OS starts killing critical processes to stay alive.

The root cause is the band steering daemon (nrd). When a modern device (especially iPhones or Samsung hardware) uses a Private/Randomized MAC address—which is enabled by default nowadays—the Deco system enters an 802.11k measurement loop every 6–15 seconds.

• Evidence in the logs: In your "Living" node, the device with MAC 74:B0:59:27:F9:7A (an Apple device) is stuck in this loop on channel 48. It generated over 8,000 log lines just by failing a measurement that never completes.

• The Result: This constant stream of errors saturates the CPU and gradually "leaks" RAM until the system collapses.

2. Why Your IoT Devices are Failing

You mentioned the system doesn't reboot but degrades. This is because as memory runs out:

• The Lua-based client manager crashes (Lua: call failed).

• The inter-node communication (TDP) breaks down, causing satellite nodes to go red or disconnect.

• The management web interface stops responding (CGI read error), which matches your observations.

In your "Bedroom" node, a Samsung device on the 6 GHz band is cycling through connect/disconnect every 10 seconds. This isn't due to interference; the Wi-Fi driver simply lacks the resources to maintain the WPA handshake.

3. The Heat Myth

While high ambient temperatures don't help with heat dissipation, the iptables and user.emerg errors in your logs point to firewall configuration corruption due to memory pressure, not thermal throttling. The fact that you experienced this with the E4 and X20 models as well confirms this is a systemic platform-level bug in the Deco codebase that has gone unpatched for years.

Potential Workarounds

Unfortunately, there is no definitive user-side fix until a firmware update is released, but you can try these "band-aids":

• Disable Private MAC Addresses: On your most frequently used devices (phones, tablets, laptops), go into the Wi-Fi settings for your home network and turn off "Private/Randomized MAC." This stops the 802.11k measurement loop for those specific clients.

• Scheduled Reboots: Even if it's annoying for your Home Assistant setup, it's currently the only way to "flush" the memory before it hits that 10,000 KB critical threshold.

• Ethernet Backhaul: If possible, wire your nodes. It slightly reduces the processing overhead on the mesh system, though it won't fix the 802.11k bug itself.

It is incredibly frustrating that a high-end system like the XE75 fails with a relatively small number of devices, but the logs are clear: this is a classic memory leak.

Have you noticed if a specific new device (like a recently updated phone) coincided with the start of these latest disconnection waves?