Figure out why it seems it measures less relays over time
As commented in https://gitlab.torproject.org/tpo/network-health/sbws/-/wikis/Bandwidth-authorities-timeline#2020-11-23, the reason might be the huge json data in state.dat to keep track of recent_measurement_attempt_count.
We have thought to remove that key/value (and to don't store that data) to confirm whether it's because of that.
#40020 (closed) would be also solved by this