IP falls off on the server, does not resolve, does not respond, where to look for a problem?

Question

Debian

Friends, recently, problems with access to sites on the VPS server have suddenly begun.
What we have: 2 IP addresses, several domains, Debian 8.
The main site sits on the main IP address. With access from the browser to it, I did not notice any problems. But there is a problem when running various scripts to automate the import-export processes. I run scripts not through CURL, but through file_get_contents. But the launch method is irrelevant.
The scripts begin to pull each other along the chain, and if the first one is launched, and it is launched a thousand times in a cycle, then at an indefinite moment it may not start. Managed to trace the error php_network_getaddresses: getaddrinfo failed: Name or service not known. Many will now say - dig into resolv.conf, see DNS. I dug already there, changed the settings. But the file itself has not changed since the server was launched, i.e. the problem is clearly not in it. Looked at etc\network\interfaces. Added a gateway to the second ip (by default it is not there), nothing changes. This whole error caught my eye because of the access to the second domain on the second IP. The site may open, it may not open, the browser hangs on an IP search. Lives its own life and much worse than the main IP. There is no IP access at this moment either. Those. DNS Trouble disappears as I understand it. There is access via SSH, if the connection is established, then it does not fall off for days via PUTTY (connection via the main IP).
DMESG looked, everything is OK there, occasionally swears at the phantom load of the kernel when executing heavy update scripts. There are no errors anywhere else.

Tell me what to dig and where to pick. It is useless to contact the hoster, I suspect hardware troubles in the hardware, maybe they have something with the equipment in the routing, but in order to indicate this they need to check their server.
Advise, maybe there are some Sh scripts that could run and identify the problem, whether it is hardware or not. Somehow it can log, hardware errors can be detected somewhere.
There seems to be enough RAM everywhere, there is no abuse from php-fpm and no muscle. Sphinx search also works without crashes, it indexes a database of 3.5 million rows without crashes. The site without problems holds 30 simultaneous requests per second. Those. Formally, everything is OK, but sometimes IPs fall off. What else to see? Installed Monitorix, this crime on the charts is invisible to me.

Answer 1

If "It is useless to contact the hoster", then my advice to you is that you need to change the hoster.
The problem can be banal when your IP hangs on someone else's virtual machine and, accordingly, traffic goes to it. You can check with the arping command at the time of its failure, but in general, hoster support should deal with this.
As a last resort, ask to change the problematic IP address.