If the Redis connection was lost, we might have missed a notification of an account being created/updated/deleted. So we fetch all accounts from DB again, then run a diff to intelligently update what we already have.
(That update process is rate-limited in the same manner as that of loading all accounts on first boot.)
We were already fetching all accounts again on reconnection, but we were ignoring them. (The diff was not implemented.)
This commit makes sure we use the latest values from DB.
b2bua: sip-bridge: account-pool: Better reflect that a full load may happen again
tests: fix: StaticAccountLoader can now return accounts multiple times
fix: ConstantRateTaskQueue: Deprecated Timer constructor
Closes FLEXISIP-475