We haven't been running workloads which create significant amounts of GCable data (i.e. old versions; txn records; abort span records), or writes larger ranges/numbers of intents (i.e. DeleteRange).

There are likely many dark corners here, some of which are already known

• storage: GC queue tuning #5252 (gc queue isn't really that well tested)
• storage: measure/optimize intent resolution #7503 (intent resolutions should piggy-back on other traffic or at least be batched up)
• stability: keep a cache of pushed transactions #627 (comment) (running into a range of intents == death)

The first two seem fairly straightforward at this point, but the last one is pretty involved. In any case, we ought to be running workloads that clearly expose those issues.