Failed to submit event to accounting, channel was blocked

[Dieterbe]

seeing these in prod sometimes.

1. we should make sure that queue fill is much less likely to happen (e.g. via batched operations instead of singular AddChunk calls)
2. we should probably make the queue blocking, and not drop messages. I haven't studied the code well enough yet, but it seems current behavior can make the bookkeeping out of touch with reality, which is bad. I rather block cache operations for several ms.

[Dieterbe]

related: should we do evictions/deletions in batches? we can optimize CCacheMetric.Del that way.

[Dieterbe]

AddRange and HitChunks implemented via #943

[Dieterbe]

we have seen some issues in prod where the cache metrics looked strange/unreliable

[Dieterbe]

the last TODO here is probably just evictions/deletions in batches. we can optimize CCacheMetric.Del

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.