dhcp server script? dhcp client script? ppp profile script?

The strange thing is that the configuration of all my routers is very similar and this only happens in one of them. And I don't understand what the "from winbox" refers to. It's really strange.

• winbox - when script is manually run from Winbox

• www - when script is manually run from Webfig (maybe also from REST API - didn't test it)

• scheduler - when script is run by scheduled event

• netwatch - when script is run by netwatch event

• dhcpclient - when script is run by DHCP Client event

• [...] - some other name of source, like events for OVPN, DHCP Server, etc. didn't test error for all ROS events...

I would recommend netinstall and configure device from scratch or export (not backup).

If you didn't run script from Windox and you have log with its source could be that device is hacked, examine if you have some job running in /system/script/job, it should be empty if there is no active CLI sessions and no running scripts/events.
I would recommend netinstall and configure device from scratch or export (not backup).

I think Mikrotik should implement a more verbose error info, since finding these errors is very time-consuming.

An error has occurred

script "mynicescript" running on device with MAC 00:ww:xx:yy:zz caused a "device not responding" error on line 42

"executing script from winbox failed, please check it manually"

So I checked my Dude setup and, after *a lot* of testing, I found that I had a function for checking the number of capsman clients that was using a call for the previous version of capsman.

Or it's up to you to well-document things you do in your network, to decrease diagnostics time

If I understand this correctly, then it's not a script, configured in ROS device (with name, etc.), but rather an in-line code being executed ... much like a command executed in CLI (so the wording "script from winbox" is a bit misleading, it comes to understanding the word "script" and my first thought is not this either).
Never the less, script execition could point at offending part if/when it dails.

:global pingResult -1
{
    :local jobID [:execute ":set pingResult [:ping count=3 8.8.8.8]"]
    :while ([:len [/system script job find where .id=$jobID]] > 0) do={
        :delay 1s
    }
}
:put $pingResult

[r***d@rtr-hq-1] > /system script run ping-check               
3

executing script from console failed, please check it manually

[root@xxx~]# snmpget -v2c -c xxx rtr-hq-1 1.3.6.1.4.1.14988.1.1.18.1.1.2.2
SNMPv2-SMI::enterprises.14988.1.1.18.1.1.2.2 = STRING: "-1"