The deployment's systemd unit passes `--log-budget`, from a new `log_budget_bytes` variable that takes a tenth of `data_volume_size_gb` unless a number is pinned: 2 GiB on the default 20 GiB volume, with the rest left to the blob store it shares. Past it a write is refused as `507 StorageFull` and the log stands where it was, and the compaction each start performs is what brings a log that filled on churn back under it. Change-Id: I2c5a8df1421e2f9dadefa5330cdfb2d421fc5c16