obsv: add loader-facing write-pressure counters (graph edge writes, dispatch capacity busy)

[EnRaiha]

Version / build tested against

origin/main @ 1ff3551

Deployment mode

Origin, single node (local)

Engine(s) involved

graph, cluster, observability

Summary

I checked what /metrics already exposes before asking for anything, and most of what a
loader needs is there: bridge_queue_depth (registry field at
nodedb/src/control/metrics/database.rs:32, writer at :109) covers queue pressure,
wal_commit_latency_p99 covers commit stalls, and qps covers generic request rate.
The registry lives at nodedb/src/control/metrics/database.rs for per-database metrics
and nodedb/src/control/metrics/system/fields.rs:16 (SystemMetrics) for system ones.
Two things I could not find:

1. A write counter per engine, for example edges written per second for graph
   (nodedb_graph_edges_written_total). Loaders size their pacing from arrival vs apply
   rate, and today they estimate it from qps, which mixes reads in.
2. A counter for dispatch capacity busy (nodedb_dispatch_capacity_busy_total). Moved to cluster: a saturated dispatch queue has no documented retryable class — clients cannot back off correctly; add one and a counter #371, which asks for the same counter.

If both already exist under names I did not find, close this with a pointer and I will
document them on my side. The capacity-busy counter is tracked in #371. This issue tracks the per-engine write counter.

What actually happened? (check all that are true)

• Acknowledged/committed data was lost, corrupted, or silently wrong
• The server crashed, hung, or failed to start
• A security or isolation boundary was crossed
• Core functionality is broken with no acceptable workaround
• A workaround exists (estimate from qps, or pace conservatively)

Proposed severity

SEV-4, observability gap. No correctness impact.

Reproducibility

Not applicable.

Last known-good version / commit (if a regression)

Not a regression.

Environment & logs

Linux x86_64, single-node origin. Catalogue read from docs/architecture.md (Per-Database
and Per-Tenant metric lists) on 2026-09-24.

Before submitting

• I searched existing issues and the docs before asking.
• This is not a security vulnerability.