There are various alternative mechanisms for effectively throttling mutator activity in favor of concurrent GC activity to avoid degeneration (aka concurrent mode failure) which would effectively convert the current stop-world degenerated gcs or full gcs into stalled allocation by mutators which may end up replicating all of the benefits of pacing but with simpler and more effective code.

GenShen by default and internal testing at Amazon for Shenandoah run with ShenandoahPacing disabled.

We want to evaluate if it would make sense to phase out (deprecate, with default off) and eventually remove ShenandoahPacing code, after checking if any customer workloads may crucially depend on this feature for good performance "at the edge".

This ticket will investigate these questions and, if feasible, deprecate and remove ShenandoahPacing flag and associated code. The implementation, in its current form, is considered a potential performance bottleneck when pacing is enabled. Appropriate performance data will be gathered to help answer these questions.