ESP8266 stops responding to AT commands after 20h of operation - how to stabilize?

I communicate with ESP via AT commands. Configuration:
AT+CWMODE=1
AT+CIPMUX=1
AT+CIPSERVER=1,80

Periodically, it queries NTP:
AT+CIPSTART=4,\"UDP\",\"193.67.79.202\",123,1112,0
AT+CIPSEND=4,48
AT+CIPCLOSE=4

This works without a problem. HTTP support too, but after about 20 hours the query on port 80 does not come (no +IPD message). A software reset, re-enabling listening (AT+CIPSERVER=1,80) solves the problem.
I checked what happens when the module "hangs up". Listening is enabled, because AT+CIPSERVER=1,80 gives an error (unfortunately I don't remember which one, next time I try I'll make a note and write it down). Trying to close port 80 with AT+CIPCLOSE=1 gives an UNLINK message.

Due to the fact that the error occurs after about 20 hours of operation, the research is going quite slowly. Has anyone encountered a similar problem?

Such a number i.e. 20h has already appeared here.
In a badly written program, the timers overflow (supposedly) precisely after this time.
I am writing from memory.

JacekCz wrote:

This number i.e. 20h has already appeared here.
In a badly written program the timers overflow (supposedly) just after this time.
I am writing from memory.

.
Except that I don't have a timer. I count the timeouts on virtual timers counting down to zero. But maybe there is something messed up in ESP?

I know that the problem can be fixed by resetting the ESP every 3 hours, for example. It starts quickly, in 10 seconds the HTTP is already running. I can try just closing and opening listening on port 80, maybe that is enough. However, I would like to find out where the problem lies. I have a PenDrive on the device so I can do logging. The question is what should I be saving? Probably everything that comes in from the ESP. Maybe I can find some sort of message about problems with port 80.

EDIT .
Today, it died after about 10 hours.

EDIT .
Now the problem again after about 10 hours. I am beginning to suspect a power supply problem. After 20 hours, the pad occurred when I powered from the 230V-USB power supply. After 10 hours when I power from the USB of the computer. Fortunately, I have the possibility to power from 7..12V and the stabiliser is in the device. I will see what the effects will be after changing the power supply.
I don't want to make logs but they would help me a lot. If the ESP does a reboot I would have that in the log.

The ESP8266 itself needs a stable power supply, and secondly it needs to be woken up periodically by loops, otherwise it likes to go into "kime"

Hubert9801 wrote:

The ESP8266 itself needs a stable power supply, and secondly it needs to be woken up periodically by loops, otherwise it likes to go "kime"

.
Box.

Quote:

.
Every once in a while, it queries NTP:

The polling is done every minute and it works. If the ESP was asleep or there were power problems I would not get a response from NTP and it would not work just listening.