I was using a common mongoDB PASS provider compose.io, and it exposes a mongo url such as:

mongodb://USER:PWD@candidate.14.mongolayer.com:11519,candidate.4.mongolayer.com:11323/app59185328?replicaSet=set-5878f8b20a2d891e52000413

(well the replicaSet part I added myself since it was missing, but they are in fact a replicaset, and the driver complains if i don't write it_

The problem is that this domains seems to be proxies. Since, if you do a rs.status() you get that the replica set is:

 c1519.candidate.14.mongolayer.com:11519 PRIMARY
 c1323.candidate.4.mongolayer.com:11323 SECONDARY
 c1160.hatch.19.mongolayer.com:11160 ARBITER
 c557.temple.44.mongolayer.com:10557 SECONDARY

So the mongodb node js driver, ends up logging:

the seedlist server was removed due to its address candidate.4.mongolayer.com:11323 not matching its ismaster.me address c1323.candidate.4.mongolayer.com:11323

So, the configuration is BAD. What happens? If you have a script that just connects to this mongourl. The driver keeps dropping and recreating the connections, eternally. And it starts slowly eating the cpu, until it has 100% load. My servers after a while were totally unresponsive, and it was quite hard to detect the issue, since there were no logs or anything. The drives logs a warning, but the default level is error.

In summary:

I know that the configuration is bad, but that doesn't mean that the driver ends up eating all the cpu (and i believe memory too, but it takes much longer)

You can try just with a script like:

const { MongoClient } = require('mongodb');
const dbUrl = "here goes the bad url";
MongoClient.connect(dbUrl, {
          db: {
            bufferMaxEntries: 0,
            w: 1
          },
          server: {
            reconnectInterval: 4000, // Restores mongo connection if somehow it is dropped.
            connectionTimeoutMS: 20000,
            socketTimeoutMS: 20000
          }
 });

Hope it helps!!!
Let me know anything i can do the help!