@Oscar Chacón this issue ended up being with the Salesforce servers running into a memory leak issue, which would then cause connections to randomly drop from one server to another.  We had to run our own packet captures and HAR browser captures to show this.  We were able to see that the requests were being sent to the SF load balancer and then the load balancer would reply with a 5xx server error.  With that being said though, that was a few years ago at this point.  They had patched that specific issue and we hadn't seen it pop back up.

FWIW - We have about 100 people in Costa Rica too and they always constantly had issues.  We actually just changed last year from Cisco with Bucher & Suter and Salesforce Omni-Channel to Salesforce Service Cloud Voice with Amazon Connect.  We still have these issues though with our CR people.  Something is wonky from region to region that causes weird latency/delays/etc over the public internet. 

I'm was still pushing my company for a local VPN in a CR DataCenter with ExpressRoutes set up to SalesForce to avoid the public internet if possible.