.gitea/workflows/go-unit.yaml

@@ -30,6 +30,17 @@ on:
       - '.gitea/workflows/go-unit.yaml'
       - '!**/*.md'
 
+# Serialised with ui-test and integration via a shared concurrency group.
+# All three run on the host runner, which shares a single workspace, so
+# running them at the same time races on workspace teardown: the
+# root-owned ui/frontend/build that ui-test's production build leaves
+# behind cannot be removed by a sibling job's (non-root) cleanup, which
+# then spuriously fails that job. One global group runs them one at a
+# time; cancel-in-progress: false queues runs instead of cancelling them.
+concurrency:
+  group: galaxy-host-runner
+  cancel-in-progress: false
+
 jobs:
   test:
     runs-on: ubuntu-latest

.gitea/workflows/integration.yaml

@@ -37,6 +37,14 @@ on:
       - '.gitea/workflows/integration.yaml'
       - '!**/*.md'
 
+# Shared concurrency group with go-unit and ui-test: the host runner uses
+# one workspace, so these jobs must not run concurrently or they race on
+# workspace teardown (see go-unit.yaml for the full rationale). Runs are
+# queued, not cancelled.
+concurrency:
+  group: galaxy-host-runner
+  cancel-in-progress: false
+
 jobs:
   integration:
     runs-on: ubuntu-latest

.gitea/workflows/ui-test.yaml

@@ -18,14 +18,16 @@ on:
 
 # Playwright launches its own `pnpm dev` on :5173, and in host-mode
 # the runner shares the host's port namespace with every other job,
-# so two parallel ui-test runs collide on EADDRINUSE. Serialise via a
-# singleton concurrency group with queueing — new runs wait their
-# turn instead of cancelling the in-progress one. cancel-in-progress
-# is explicitly false because Gitea has shown spurious self-cancel
-# behaviour under cancel-in-progress: true even when no other run
-# shares the group.
+# so two parallel ui-test runs collide on EADDRINUSE. Serialise via the
+# shared `galaxy-host-runner` concurrency group (also used by go-unit and
+# integration) with queueing — new runs wait their turn instead of
+# cancelling the in-progress one. The shared group additionally avoids the
+# host-workspace teardown race those jobs hit (see go-unit.yaml). New runs
+# wait their turn; cancel-in-progress is explicitly false because Gitea has
+# shown spurious self-cancel behaviour under cancel-in-progress: true even
+# when no other run shares the group.
 concurrency:
-  group: ui-test-singleton
+  group: galaxy-host-runner
   cancel-in-progress: false
 
 jobs: