v0.13.0

[0.13.0] — 2026-07-29

Added — spans over the phases nobody could see, and six report fields to match

Internal tracing instrumentation over the run path, and six additive
RunReport fields: boot_ms, teardown_ms, copy_out_ms,
snapshot_build_ms, and the split of the existing commit_ms into
commit_hash_ms + commit_copy_ms. Each field is absent when its phase did
not happen, so every existing consumer sees the exact JSON it always did.
resume_ms and commit_ms themselves are unchanged.

What the numbers showed once they existed:

• Teardown costs ~140 ms on every run — 27% of a 509 ms cold run, second
  only to the kernel boot — and had been folded into a "boot" figure nobody
  computed directly. Of ~470 ms non-exec time the kernel boot proper is
  ~210 ms: the number an operator would have read as boot was about twice the
  real thing.
• Copy-out moves ~52 MB/s over the base64-in-JSON vsock path, and runs
  after the timeout_s budget has stopped protecting the caller.
• commit_ms is now split, not diagnosed. The hash and copy passes are
  separately visible; on real commits they are the same order of magnitude,
  with end-to-end time dominated by writeback of the just-written scratch (see
  0.12.5 below for the measured numbers and the buffered-read fix they led to).

What this does NOT buy: it is instrumentation, not telemetry. There is no
exporter, no network path, and no opentelemetry dependency in any build.
With RUST_LOG unset the binaries install no subscriber and write zero bytes;
RUST_LOG=isopod=debug prints the spans to stderr and nowhere else. Span
attributes pass through a sealed Attr type whose only string carriers are
&'static str and the host-minted vm_id, and guest-influenced magnitudes
appear only as log2 buckets. exec_ms still folds vsock output streaming into
compute time. The two spans written blind while the host's taps were down —
isopod.run.snapshot_ensure and the warm isopod.run.resume — have since run
live: a first-warm run reported snapshot_build_ms: 4238 / resume_ms: 72
with both spans on stderr, and a second run of the same shape reported
resume_ms: 61 with snapshot_build_ms correctly absent.