INTERNAL_ERROR is an http2 error which literally means the server had an unexpected internal problem. This is very likely some problem on the server or in the way k6 interacts with the server. There should be server logs for this. If you have any kind of API gateways you should probably look into them as well.
If this happens only under load I would argue it is the server being overloaded :). Try with smaller amounts of VUs and see if it disappears and go up until it starts again.
Some issues in the golang project (k6 is written in golang) has shown that possible problems are writing/reading too slow which is not a thing (that I think) k6 can do. Are you having particularly big bodies? Also a server side project has fixed that by reading the request body fully on the server. K6 does read fully the response bodies even if you wouldn’t need them for exactly this reason.
There is 36759, but both me and another colleague have been running huge tests recently and I don’t remember ever seeing INTERNAL_ERROR so I still would argue it is with the server and the golang version probably just changed how much traffic golang send and as that how much the server is under stress