This will allow updating metadata of repositories known to be inactive in terms of source-code, but may have changes in metadata (eg. watchers, stargazers, ...)
We decided to add recurring fetches, so it will take care both of backfilling now, and visiting from time to time in the future. We're going to assume 3 months for now, as it seems reasonable to not exhaust rate limits.
The plan:
add metrics for token usage, as we need that to make decisions
add a swh-scheduler component (CLI called from a cron?) that reads from the DB and fills a Kafka topic
a journal client that reads from the topic and fetches the metadata
For now, we are only going to use a single topic, which means the whole thing will block when we exhaust a token's rate-limit; but the aim is to not hit rate-limits at all. We can revisit the question later, anyway.