I was wondering whether we really need to fetch the deviceID on every Tunnelbroker connection, so I spent some time thinking about what scenarios might lead to a reconnect I came up with two scenarios:

1. If the deviceID changes for some reason
2. If a connection error forces a reconnection

Since we'll need some code to handle the first case anyways (forcing a disconnect and reconnecting to Tunnelbroker with the new device ID), I wonder if we can avoid having to fetchOlmAccount on every connection. Here's what I'm thinking:

1. We have some function like createAndMaintainTunnelbrokerSocket that takes deviceID as input. This function internally handles reconnecting in the case of an error or a disconnect
2. When the keyserver starts, we fetch the deviceID (just once) and call createAndMaintainTunnelbrokerSocket
3. Separately, we'll add code to every place that might change the deviceID in the database, and make sure it forces a reconnection with the updated deviceID
   • We'll need some way to kill the existing createAndMaintainTunnelbrokerSocket instance... either the function can return some sort of handle, or we can use a class instead of a function

What do you think?

Promise.all isn't doing anything here

no, you're right. Got confused

Yea, sounds good, I'll refactor this.

I intended to extend this with grabbing the user_id and access token as well.

I'll refactor this to have the local fetching done before attempting the session.