diff --git a/_toolbox/PR_DESCRIPTION.md b/_toolbox/PR_DESCRIPTION.md
index c274d458..6d3ee776 100644
--- a/_toolbox/PR_DESCRIPTION.md
+++ b/_toolbox/PR_DESCRIPTION.md
@@ -72,7 +72,7 @@ I tried contributing some of these fixes back to `microsoft/dbt-fabric` first, b
 
 - [#01 — Six `__exit__` methods return `True` (silent exception swallowing)](https://github.com/sdebruyn/dbt-fabric/blob/to-toolbox/_toolbox/upstream_issues/dbt-fabricspark/01-exit-methods-return-true-swallow-exceptions.md)
 - [microsoft/dbt-fabricspark#192 — Hardcoded `expires_on = 1845972874` (year 2028) in `int_tests` auth path bypasses all token refresh](https://github.com/microsoft/dbt-fabricspark/issues/192)
-- [#04 — `_getLivySQL`: `re.DOTALL` passed as positional `count` caps comment-stripping at 16](https://github.com/sdebruyn/dbt-fabric/blob/to-toolbox/_toolbox/upstream_issues/dbt-fabricspark/04-getlivysql-regex-bug.md)
+- [microsoft/dbt-fabricspark#194 — `_getLivySQL`: `re.DOTALL` passed as positional `count` caps comment-stripping at 16](https://github.com/microsoft/dbt-fabricspark/issues/194) (draft fix: [microsoft/dbt-fabricspark#195](https://github.com/microsoft/dbt-fabricspark/pull/195))
 - [#06 — Livy session cleanup bypasses dbt's `close()` lifecycle and uses `atexit`](https://github.com/sdebruyn/dbt-fabric/blob/to-toolbox/_toolbox/upstream_issues/dbt-fabricspark/06-atexit-handlers-leak-livy-sessions.md)
 - [#07 — Dead Thrift exception handler from dbt-spark ancestry](https://github.com/sdebruyn/dbt-fabric/blob/to-toolbox/_toolbox/upstream_issues/dbt-fabricspark/07-thrift-dead-code.md)
 - [#08 — Proposal: inherit from `dbt-spark` instead of being a standalone `SQLAdapter`](https://github.com/sdebruyn/dbt-fabric/blob/to-toolbox/_toolbox/upstream_issues/dbt-fabricspark/08-inherit-from-dbt-spark.md)
diff --git a/_toolbox/upstream_issues/INDEX.md b/_toolbox/upstream_issues/INDEX.md
index 5f25da05..c1edc220 100644
--- a/_toolbox/upstream_issues/INDEX.md
+++ b/_toolbox/upstream_issues/INDEX.md
@@ -43,7 +43,7 @@ Each issue body carries a `> [ ] Validated by maintainer` checkbox at the top. S
 
 - [x] 01 — Six `__exit__` methods return `True`
 - [x] 02 — Hardcoded 2028 token expiry → filed as [microsoft/dbt-fabricspark#192](https://github.com/microsoft/dbt-fabricspark/issues/192)
-- [x] 04 — `_getLivySQL` regex bug
+- [x] 04 — `_getLivySQL` regex bug → filed as [microsoft/dbt-fabricspark#194](https://github.com/microsoft/dbt-fabricspark/issues/194), draft PR [microsoft/dbt-fabricspark#195](https://github.com/microsoft/dbt-fabricspark/pull/195)
 - [x] 06 — Livy cleanup bypasses dbt's `close()` lifecycle
 - [x] 07 — Thrift dead code from dbt-spark ancestry
 - [x] 08 — Proposal: inherit from `dbt-spark`
@@ -111,7 +111,7 @@ These are debatable and may invite long discussion. File only if the second batc
 |---|---|---|---|
 | 01 | Six `__exit__` methods return `True` — silent exception swallowing | high | bug |
 | 02 | [Hardcoded 2028 token expiry bypasses refresh logic](https://github.com/microsoft/dbt-fabricspark/issues/192) | high | bug, security |
-| 04 | `_getLivySQL` regex bug: `re.DOTALL` passed as positional `count` | medium | bug |
+| 04 | [`_getLivySQL` regex bug: `re.DOTALL` passed as positional `count`](https://github.com/microsoft/dbt-fabricspark/issues/194) | medium | bug |
 | 06 | `atexit` handlers leak Livy sessions on hard kill / OOM | medium | bug |
 | 07 | Thrift dead code from dbt-spark ancestry | low | tech-debt |
 | 09 | `botocore`/`boto3` DEBUG logging at import time | low | tech-debt |
diff --git a/_toolbox/upstream_issues/dbt-fabricspark/04-getlivysql-regex-bug.md b/_toolbox/upstream_issues/dbt-fabricspark/04-getlivysql-regex-bug.md
deleted file mode 100644
index e1827da9..00000000
--- a/_toolbox/upstream_issues/dbt-fabricspark/04-getlivysql-regex-bug.md
+++ /dev/null
@@ -1,45 +0,0 @@
-# `_getLivySQL`: `re.DOTALL` passed as positional `count` arg — comment-stripping silently capped at 16 replacements per file
-
-**Repo:** `microsoft/dbt-fabricspark`
-**Labels (suggested):** `bug`, `priority/medium`
-
-> [x] **Validated by maintainer** — code refs, line numbers, and claims confirmed against upstream HEAD
-
-> **Internal note (strip before filing):** Submittable as a PR — swap `re.DOTALL` from positional to `flags=re.DOTALL` in both copies; extracting a shared `_strip_sql_comments(sql)` helper is a small bonus to keep the bug from reoccurring. Consider opening with the issue *and* a draft PR linked from it.
-
-## Summary
-
-`_getLivySQL` passes `re.DOTALL` as the positional `count` argument to `re.sub` instead of as `flags=re.DOTALL`. `re.DOTALL` is an integer enum value (16). `re.sub`'s positional signature is `re.sub(pattern, repl, string, count=0, flags=0)`. Passing `re.DOTALL` positionally sets `count=16`, capping comment-stripping to at most 16 replacements per file.
-
-## Evidence (HEAD [`d315a56`](https://github.com/microsoft/dbt-fabricspark/tree/d315a56))
-
-`_getLivySQL` is duplicated verbatim between [`src/dbt/adapters/fabricspark/singleton_livy.py#L488`](https://github.com/microsoft/dbt-fabricspark/blob/d315a56/src/dbt/adapters/fabricspark/singleton_livy.py#L488) and [`src/dbt/adapters/fabricspark/concurrent_livy.py#L555`](https://github.com/microsoft/dbt-fabricspark/blob/d315a56/src/dbt/adapters/fabricspark/concurrent_livy.py#L555). Both copies call `re.sub(pattern, '', sql, re.DOTALL)` (positional) when they meant `re.sub(pattern, '', sql, flags=re.DOTALL)`.
-
-## User impact
-
-- Model SQL files with more than 16 `/* */` comment blocks have some comments left in the SQL submitted to Spark.
-- Most of the time Spark's parser shrugs them off, but commented-out SQL can introduce surprises — particularly when the leftover comment contains characters that interact with the Livy/Spark parser, or when the comment contains SQL keywords that the parser must lexically skip past.
-- Without `re.DOTALL` set, comment blocks containing newlines are not matched at all by the multi-line comment pattern — so the bug is doubly broken: count is set to 16, and the DOTALL flag that the code intended to set is not set.
-
-## Reproduction
-
-```python
-import re
-pattern = r"/\*.*?\*/"
-sql = ("/* a */\n" * 20) + "select 1"
-# Buggy call:
-result_buggy = re.sub(pattern, "", sql, re.DOTALL)
-# count=16 is set; flags=0 (no DOTALL); .*? does not cross newlines anyway
-# but the count cap kicks in regardless.
-```
-
-## Suggested fix
-
-Use keyword args:
-
-```python
-sql = re.sub(pattern, "", sql, flags=re.DOTALL)
-```
-
-Fix both copies ([`singleton_livy.py#L488`](https://github.com/microsoft/dbt-fabricspark/blob/d315a56/src/dbt/adapters/fabricspark/singleton_livy.py#L488) and [`concurrent_livy.py#L555`](https://github.com/microsoft/dbt-fabricspark/blob/d315a56/src/dbt/adapters/fabricspark/concurrent_livy.py#L555)). Better: extract a single `_strip_sql_comments(sql)` helper and call it from both files so the bug can't reoccur in only one place.
-