fix(chart): the ingestor tag fallback is per-environment, not always prod's

[LukasWodka]

Summary

tracebloc.ingestorTag's last-resort literal was a bare "0.8" for every
environment. On the --reuse-values replay it exists for — a release predating
channelTags, where no chart default is adopted — a dev or staging edge
spawned the PROD ingestor line.

That inverts the whole point of backend#1360: dev/stg channels exist so an
ingestor change can be validated on a real edge without a prod release. An
edge silently validating prod's image reports on the wrong artifact.

It is no longer merely a wrong tag. The prod float has moved past the
ordering ceiling documented at values.yamlprodDigest: the 0.8 line no
longer carries the ingestor's edgeuserDB_USER default that
data-ingestors#468 removed (backend#1853). serviceDbAccountsByEnv supplies
DB_USER on dev/stg, so those two survive it — but the coupling is accidental,
and the same literal is where an out-of-vocabulary CLIENT_ENV lands, where
serviceDbAccountsByEnv misses too and nothing supplies DB_USER. That is
backend#1752 reconstructed from a typo.

Fallback is now keyed on the resolved environment (dev→dev, stg→stg,
prod→0.8), so the documented aliases reach it too.

Verified on the real chart — input reaching output, with a control

Rendered against client/ci/bm-values.yaml, helm v4.1.1.

Before (origin/develop @ ead58c6) — channelTags: null, i.e. the replay
this branch serves:

channelTags:null CLIENT_ENV=dev -> INGESTOR_IMAGE_TAG value: "0.8" <-- prod line on a dev edge
channelTags:null CLIENT_ENV=stg -> INGESTOR_IMAGE_TAG value: "0.8" <-- prod line on a stg edge
channelTags:null CLIENT_ENV=prod -> INGESTOR_IMAGE_TAG value: "0.8"

Control proving the opposite input differs (same command, channelTags present):

channelTags present, CLIENT_ENV=dev -> INGESTOR_IMAGE_TAG value: "dev"

After (this branch):

dev channelTags:null -> value:"dev"
stg channelTags:null -> value:"stg"
prod channelTags:null -> value:"0.8" <-- control: prod unchanged
development channelTags:null -> value:"dev" <-- alias reaches the fallback
staging channelTags:null -> value:"stg"
production channelTags:null -> value:"0.8"
produktion channelTags:null -> value:"0.8" <-- unknown still renders, unchanged

The duplication is now documented in both directions

The prod literal is a second copy of values.yamlchannelTags.prod and
cannot read the first: --reuse-values (unlike --reset-then-reuse-values)
does not adopt new chart defaults, so a values lookup is nil on exactly the
releases this branch serves. So each copy now points at the other, and the suite
pins both — a bump that touches only one fails CI instead of drifting.

Tests

An existing case asserted 0.8 for a dev edge with channelTags absent —
the bug written down as an expectation. Replaced by four cases plus a prod
control, so "dev falls back to dev" cannot pass by merely echoing the
environment.

make check (lint + drift guards + helm lint, all 4 platform value files):

1 chart(s) linted, 0 chart(s) failed (x5: default, aks, bm, eks, oc)
==> check: green (bats + helm unit tests are in 'make check-all')

make helm-template — all four platforms render:

=== Rendering aks ===
=== Rendering bm ===
=== Rendering eks ===
=== Rendering oc ===

make helm-unittest:

Charts: 1 passed, 1 total
Test Suites: 30 passed, 30 total
Tests: 388 passed, 388 total
Time: 2.04795475s

Mutation-checked — the new cases actually catch the bug rather than
describing it:

MUTATION 1: restore the old bare "0.8" literal
Tests: 3 failed, 385 passed, 388 total
MUTATION 2: fallback echoes the resolved env (naive fix — breaks prod)
Tests: 2 failed, 386 passed, 388 total <-- the prod control earns its place
restored -> Tests: 388 passed, 388 total

scripts/gen-manifest.sh --check passes unchanged: the manifest covers
scripts/** only, and this PR touches no file it hashes.

Type

• Bug fix (non-breaking)

Notes for the reviewer

• Not a breaking change. Prod behaviour is byte-identical; only dev/stg
  edges whose channelTags are absent change, and they change from "prod's
  image" to "their own channel", which is what the schema already promises.
• Chart version bumped 1.9.34 → 1.9.35 (values.yaml + templates/** are
  chart content per scripts/chart-version-guard.sh).
• Sibling PR closes the CLIENT_ENV / channelTags vocabulary that lets an
  edge reach this fallback by typo in the first place. That one is breaking and
  is flagged for your decision; this one is independent and safe on its own.

---

Note

Medium Risk
Changes which ingestor image tag dev/stg edges use when channelTags is absent; prod behavior is unchanged but ingestion spawn paths are operationally sensitive.

Overview
tracebloc.ingestorTag no longer uses a single last-resort "0.8" when images.ingestor.channelTags is missing (typical --reuse-values upgrades). It now maps resolvedCLIENT_ENV to dev→dev, stg→stg, prod→0.8, with unknown envs still defaulting to 0.8.

That stops dev/staging from spawning the prod ingestor line when channel tags were never stored—restoring backend#1360’s intent and avoiding accidental prod-image validation on non-prod edges (including DB_USER coupling with the 0.8 line).

Docs and tests:_helpers.tpl and values.yaml document the duplicated prod literal vs channelTags.prod; ingestor_channel_tag_test.yaml replaces the old “dev gets 0.8” expectation with per-env fallback, prod control, alias, and sync checks. Chart version 1.9.37 → 1.9.38.

^{Reviewed by Cursor Bugbot for commit 59d26e0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[saadqbal]

Clean fix. Fallback is now keyed on the resolved env via get $fallbacks $clientEnv | default "0.8", so dev→dev, stg→stg, prod→0.8, and an out-of-vocab env still lands on 0.8 (no empty render). Runs through tracebloc.clientEnv so aliases normalize before the key lookup, and the whole path stays nil-guarded for the --reuse-values replay (channelTags absent) this exists for. The prod-control + values-vs-literal pin tests are the right guards against "echo the env" and silent drift. prod unchanged (still digest-pinned). Chart bumped 1.9.34→1.9.35. helm-unittest green locally (17/17). 👍

[saadqbal]

Re-reviewed at the merge head — the develop merge is clean. Chart.yaml conflict correctly resolved to 1.9.38 (one above develop's 1.9.37), and the helper + test auto-merged coherently with #695's vocabulary close. Per-env fallback and the values-vs-literal drift guard intact, full helm-unittest 394/394 green. 👍