Default Artemis GC Settings are not adequate for production systems

During testing customer (IG Index) are seeing high amounts of GC, causing significant latency spikes unto 1 minute, quite periodically.
We are running 2GB heap (xms, xmx set to this) with parallel gc.
We would like to look into tuning this to reduce the GC spikes, and general latency overhead.
Where are you experiencing the behavior? What environment?
We don't expect a stw GC >100ms, and we expect at most 2 stw per day.
When does the behavior occur? Frequently? Repeatedly? At certain times?
Frequently

Customer has already been working with Clebert via open support ticket:

https://access.redhat.com/support/cases/#/case/01782523

Reproducing this in Jira as requested by Andy Taylor