As mentioned by ballin on IRC (https://chatlogs.metabrainz.org/brainzbot/metabrainz/msg/4716182/): stopping and starting Docker while fetching dumps makes to loose all the progress made so far. This is because the script simply remove all existing files before downloading either search indexes or database dumps, see https://github.com/metabrainz/musicbrainz-docker/blob/9fb99ecfde4d821ae2887ceaae51bc736d96fd21/build/musicbrainz-dev/scripts/fetch-dump.sh#L89-L109.