Fix flaky prometheus_updater_spec by cleaning up PROCESS_TYPE env var

[joyvuu-dave]

Fix flaky prometheus_updater_spec caused by PROCESS_TYPE env var leaking between tests

Problem

The PrometheusUpdater spec has been flaky since PR #4749 (commit 9c9a9d51d, merged Jan 16 2026), which introduced ExecutionContext and made metric registration conditional on ENV['PROCESS_TYPE']. CI runs frequently fail with errors like:

NoMethodError:
  undefined method `set' for nil
# ./lib/cloud_controller/metrics/prometheus_updater.rb:131:in `update_gauge_metric'

Root Cause

Three spec files set ENV['PROCESS_TYPE'] (via set_process_type_env) but never restore it:

• connection_metrics_spec.rb — sets it to cc-worker and puma_worker. The cc-worker value is the direct cause of the flakiness: it maps to CC_WORKER, which only registers DB_CONNECTION_POOL_METRICS, DELAYED_JOB_METRICS, and VITAL_METRICS. The puma_worker value happens to be harmless today (API_PUMA_WORKER registers all metrics), but it is still pollution.
• puma_runner_spec.rb — sets it to puma_worker (via before_worker_boot callback). Also harmless today for the same reason, but still pollution.
• runner_spec.rb — sets it to main (via Runner#initialize), which maps to API_PUMA_MAIN — the same context the test environment defaults to. Harmless today, but still pollution.

With randomized test ordering, when connection_metrics_spec.rb's cc-worker context runs before prometheus_updater_spec, the leaked value changes the behavior of ExecutionContext.from_process_type_env:

• Normal state: PROCESS_TYPE is unset → falls back to CC_TEST=true check → returns API_PUMA_MAIN → registers all metrics
• Polluted state: PROCESS_TYPE=cc-worker → returns CC_WORKER → registers only a subset of metrics → metrics like :cc_deployments_in_progress_total are never registered → @registry.get(...) returns nil

Note: connection_metrics_spec.rb was already setting ENV['PROCESS_TYPE'] without cleanup before PR #4749, but it didn't matter then because PrometheusUpdater registered all metrics unconditionally. PR #4749 made registration conditional, which turned the pre-existing env var pollution into a flaky test.

Fix

Add around blocks to all three polluting specs that save and restore ENV['PROCESS_TYPE']:

around do |example|
  original_process_type = ENV.fetch('PROCESS_TYPE', nil)
  example.run
ensure
  if original_process_type.nil?
    ENV.delete('PROCESS_TYPE')
  else
    ENV['PROCESS_TYPE'] = original_process_type
  end
end

Verification

Reproducible with --seed 8:

bundle exec rspec --seed 8 \
  spec/unit/lib/sequel/extensions/connection_metrics_spec.rb \
  spec/unit/lib/cloud_controller/runners/puma_runner_spec.rb \
  spec/unit/lib/cloud_controller/runner_spec.rb \
  spec/unit/lib/cloud_controller/metrics/prometheus_updater_spec.rb

• Without fix (on main): 74 examples, 11 failures — the exact same 11 prometheus_updater_spec failures seen in CI
• With fix: 74 examples, 0 failures

Additionally tested with seeds 1–7, 9, 10, 11111, 22222, 33333, 44444, 55555, 66666, 12345, and 67890 — all pass.

• I have reviewed the contributing guide

• I have viewed, signed, and submitted the Contributor License Agreement

• I have made this pull request to the main branch

• I have run all the unit tests using bundle exec rake

• I have run CF Acceptance Tests