Sidekiq use connection pool internally, with short timeout (1s by default)
On heavy load of workflow creation (>1000 in same time), this pool is quickly
filled and job creation failed with a timeout after waiting for pool allocation
Seems this is not a trouble for Sidekiq base worker creation but only for
workflow/job here
Sidekiq doc seems saying the connection pool is sized for internal purpose only
(size = 2 + concurrency)
So move to own connection pool with longer timeout to workaround the deadlock
and to avoid any possible conflict with Sidekiq
aeris