After consulting with several people on several departments, we have scheduled a maintenance window (downtime) for dbstore1002 (alias analytics-store, and s2 *to* s7-analytics-slave) for Thursday 21 July between14:00-15:00 UTC [0].
The downtime is expected to be of only 5-10 minutes, but in case something goes wrong, we will reserve the entire hour. During the maintenance, this host will stop replication from eventlogging and the production shards, kill all ongoing queries and stop responding to new ones. User databases on this host will be unavailable, too.
The recommendation is to use db1047 / analytics-slave / s1-analytics-slave, which also has an up-to-date version of eventlogging and shards s1 and s2 for the following 3 days to avoid service interruption. I see some crons running on dbstore1002, contact me privately if you want to change them and require help.
I will send an update when the maintenance is over to confirm normal work can resume.