We'll also want to document here (and possibly figure out first, if not currently known) why it is that we specify a local replica as "master" in wmf-config for secondary DCs.

MediaWiki itself and the rdmbs library do not have any concept for a local/stand-by master. Right now the wmf-config is using a local replica for this field, which to my knowledge was a safety precaution to use until a time we have more confidence in our switchover procedures – to be phased out at some point before we enable multi-dc/active-active traffic.

So as written, this task is to either:

1. Update wmf-config to configure MW the way MW currently supports being configured: It must be configured with the real master, not a local replica. The reason this isn't an issue right now is because the secondary config is essentially never used, because we're not multi-dc/active-active yet.
2. Document a reason for why MW would need to know about a local stand-by master, add a feature to MW for configuring that separately from the current master config, update wmf-config to set the real master as the master, and then use this new config for the stand-by master.

In T224419#5215482, @Krinkle wrote:

We'll also want to document here (and possibly figure out first, if not currently known) why it is that we specify a local replica as "master" in wmf-config for secondary DCs.

I believe because it is technically a master (or an intermediate read-only master if you will, maybe we can call it DC master) for the rest of replicas in that given DC.