Unpredictable behavior of registrator and consul - docker

Question

I have very simple docker-compose config:
version: '3.5'
services:
consul:
image: consul:latest
hostname: "consul"
command: "consul agent -server -bootstrap-expect 1 -client=0.0.0.0 -ui -data-dir=/tmp"
environment:
SERVICE_53_IGNORE: 'true'
SERVICE_8301_IGNORE: 'true'
SERVICE_8302_IGNORE: 'true'
SERVICE_8600_IGNORE: 'true'
SERVICE_8300_IGNORE: 'true'
SERVICE_8400_IGNORE: 'true'
SERVICE_8500_IGNORE: 'true'
ports:
- 8300:8300
- 8400:8400
- 8500:8500
- 8600:8600/udp
networks:
- backend
registrator:
command: -internal consul://consul:8500
image: gliderlabs/registrator:master
depends_on:
- consul
links:
- consul
volumes:
- /var/run/docker.sock:/tmp/docker.sock
networks:
- backend
image_tagger:
build: image_tagger
image: image_tagger:latest
ports:
- 8000
networks:
- backend
mongo:
image: mongo
command: [--auth]
ports:
- "27017:27017"
restart: always
networks:
- backend
volumes:
- /mnt/data/mongo-data:/data/db
environment:
MONGO_INITDB_ROOT_USERNAME: root
MONGO_INITDB_ROOT_PASSWORD: qwerty
postgres:
image: postgres:11.1
# ports:
# - "5432:5432"
networks:
- backend
volumes:
- ./postgres-data:/var/lib/postgresql/data
- ./scripts:/docker-entrypoint-initdb.d
restart: always
environment:
POSTGRES_PASSWORD: qwerty
POSTGRES_DB: ttt
SERVICE_5432_NAME: postgres
SERVICE_5432_ID: postgres
networks:
backend:
name: backend
(and some other services)
Also I configured dnsmasq on host to access containers by internal name.
I spent couple of days, but still not able to make it stable:
1. Very often some services are just not get registered by registrator (sometimes I get 5 out of 15).
2. Very often containers are registered with wrong ip address. So in container info I have one address(correct), in consul - another (incorrect). And when I want to reach some service by address like myservice.service.consul I end up at wrong container.
3. Sometimes resolution fails at all even when containers are registered with correct ip.
Do I have some mistakes in config?

Answer 1

So, at least for now I was able to fix this by passing -resync 15 param to registrator. Not sure if it's correct solution, but it works.