Backport #104296 to Antalya 25.8 - Disable additional_table_filters with parallel replicas and analyzer and w/o query plan serialization

[mkmkme]

Same as #2065, but for Antalya.

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Disable additional_table_filters with parallel replicas and analyzer and w/o query plan serialization (ClickHouse#104296 by @azat)

Documentation entry for user-facing changes

...

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Performance tests
• Aarch64 tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All Regression
• Disable CI Cache

Regression jobs to run:

• Fast suites (mostly <1h)
• Aggregate Functions (2h)
• Alter (1.5h)
• Benchmark (30m)
• ClickHouse Keeper (1h)
• Iceberg (2h)
• LDAP (1h)
• OAuth (5m)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [e1753f8]

[mkmkme]

@blau-ai

[blau-ai]

CI triage — #2109

Verdict: 4 failing checks, 0 PR-caused. All red is either an unrelated test, the external regression framework, or praktika/infra. The change here is scoped to src/Planner/Planner.cpp + the new 04207_pr_additional_filters test, and that new test passed — no PR-side fix is needed. Recommendation: re-run the flaky stateless job; the rest are environment/framework issues outside this PR.

Check	Class	PR-caused?
Stateless tests (amd_debug, AsyncInsert, s3 storage, parallel)	flaky / unrelated test	❌ No
Regression x86 iceberg_2	test-framework bug (ImportError)	❌ No
Docker keeper image	praktika pre-run infra	❌ No
Docker server image	praktika pre-run infra	❌ No
PR (aggregate gate)	rolls up the above	❌ No

---

1. Stateless tests (amd_debug, AsyncInsert, s3 storage, parallel) — Failed: 1

The single failing test is 03572_export_merge_tree_part_to_object_storage, not this PR's test:

2026-07-23 14:21:11 03572_export_merge_tree_part_to_object_storage:   [ FAIL ] 66.07 sec.

(The two [ BROKEN ] entries — 03441_deltalake_clickhouse_public_datasets / _virtual_columns — are known-broken deltalake tests counted separately, also unrelated.)

03572 is an S3 object-storage export test (tests/queries/0_stateless/03572_export_merge_tree_part_to_object_storage.sh, exercises MergeTree → object-storage part export). It has zero relationship to the parallel-replicas / additional_table_filters planner change and is not touched by this PR. S3-backed .sh tests in this config are a common flakiness source. This PR's own test 04207_pr_additional_filters did not appear in the fail/broken list and passed.
→ Next step: safe to re-run this job. If 03572 keeps failing across re-runs it's a pre-existing flake on antalya-25.8, not this PR.

2. Regression x86 iceberg_2 — 1 module (1 errored)

Not a ClickHouse failure — an import error in the external clickhouse-regression suite:

✘ [ Error ] '/iceberg/iceberg table engine' (18m 15s)
✘ [ Error ] '/iceberg' (1h 23m)
ImportError: cannot import name 'check_if_antalya_post_26_3_10_20001'
             from 'helpers.common' (.../iceberg/../helpers/common.py)

The regression checkout is missing a helper (check_if_antalya_post_26_3_10_20001) that the iceberg tests import — a version-skew bug in the regression repo, independent of any ClickHouse source change. 28/29 iceberg features passed; only the collection that imports that helper errored. iceberg_1 and all other x86 regression suites (parquet, s3_export_partition, swarms, …) are green.
→ Next step: nothing to do in this PR. Needs the clickhouse-regression suite pinned to a revision that has the helper (infra/CI side).

3 & 4. Docker keeper image / Docker server image — error

Both self-report:

Failed to do a job pre-run step, it is praktika bug or misconfiguration

Pre-run/setup failure in the CI orchestrator (praktika), before any of this PR's code is exercised. Pure infra.
→ Next step: safe to re-run.

PR gate

Aggregate status — red only because it rolls up the four items above. No independent failure.

---

Bottom line: this backport is healthy. The diff is minimal and correct in shape — it rejects additional_table_filters + parallel replicas on the analyzer path (serverError SUPPORT_IS_DISABLED for enable_parallel_replicas=2, silent fallback for =1), matching upstream ClickHouse#104296 adapted for 25.8 where PR followers always get the rewritten AST. Its dedicated test passed; every red check is unrelated flake/framework/infra. Re-running should clear the stateless and Docker jobs; the iceberg ImportError needs a regression-suite fix and shouldn't block the merge.

_{🤖 Automated CI triage by @blau-ai. I can't build/run ClickHouse in-container, so classifications are evidence-based from CI logs.}