Choose timeout depending on

user (priority assigned by owner of share or DB)

DB (size and complexity and location of user)

product used to connect to MD

adjust timeout dynamically (based on testing and experience gained)

influence connection pool timeouts and reconnects

vary depending on DB or share

Perspective of GoodData BI platform already integrated with MotherDuck:

We connect to MD from a Kotlin microservice using Hikari.

Hikari supports the concept of connection timeout.

It supports even more(timeout related props):

val expireAfterAccess: Long = 1200,

override val idleConnectionTimeout: Long = 10000,

override val maxConnectionLifeTime: Long = 30000,

override val minConnectionsIdle: Int = 0,

override val connectionTimeout: Long = 60000,

override val initializationFailTimeout: Long = 1,

It is possible to specify the above properties per connection (pool).

We obviously create separate connection for each combination of [DB, schema, user(token)].