diff --git a/sqlmesh/core/dialect.py b/sqlmesh/core/dialect.py
index 67918b6d14..63e6608776 100644
--- a/sqlmesh/core/dialect.py
+++ b/sqlmesh/core/dialect.py
@@ -782,6 +782,11 @@ def _parse_interval_span(self: Parser, this: exp.Expr) -> exp.Interval:
 
 def _override(klass: t.Type[Tokenizer | Parser], func: t.Callable) -> None:
     name = func.__name__
+    if getattr(klass, name, None) is func:
+        # Already installed, so re-saving it under f"_{name}" would make the wrapper delegate to
+        # itself and recurse infinitely, as well as lose sqlglot's original implementation
+        return
+
     setattr(klass, f"_{name}", getattr(klass, name))
     setattr(klass, name, func)
 
@@ -1170,11 +1175,12 @@ def extend_sqlglot() -> None:
                 MacroDef,
             )
 
-        generator.UNWRAPPED_INTERVAL_VALUES = (
-            *generator.UNWRAPPED_INTERVAL_VALUES,
-            MacroStrReplace,
-            MacroVar,
-        )
+        if MacroVar not in generator.UNWRAPPED_INTERVAL_VALUES:
+            generator.UNWRAPPED_INTERVAL_VALUES = (
+                *generator.UNWRAPPED_INTERVAL_VALUES,
+                MacroStrReplace,
+                MacroVar,
+            )
 
     _override(Parser, _parse_select)
     _override(Parser, _parse_statement)
diff --git a/tests/core/test_dialect.py b/tests/core/test_dialect.py
index e2f1daba3d..f8a1bd48b9 100644
--- a/tests/core/test_dialect.py
+++ b/tests/core/test_dialect.py
@@ -1185,3 +1185,29 @@ def test_pipe_syntax():
         ast.sql("bigquery")
         == "SELECT * FROM (WITH __tmp1 AS (SELECT id FROM t2) SELECT * FROM __tmp1)"
     )
+
+
+def test_extend_sqlglot_is_idempotent():
+    # extend_sqlglot() patches sqlglot's global classes and already ran at import time, so a
+    # second application must not re-save the installed wrappers as the "original" methods
+    from sqlglot.generator import Generator
+    from sqlglot.parser import Parser
+    from sqlglot.tokens import Tokenizer
+
+    patched_parse_types = Parser._parse_types
+    original_parse_types = Parser.__dict__["__parse_types"]
+    unwrapped_interval_values = Generator.UNWRAPPED_INTERVAL_VALUES
+    transforms = dict(Generator.TRANSFORMS)
+    var_single_tokens = set(Tokenizer.VAR_SINGLE_TOKENS)
+
+    d.extend_sqlglot()
+
+    assert Parser._parse_types is patched_parse_types
+    # the wrapper delegates to this, so saving the wrapper here is what causes infinite recursion
+    assert Parser.__dict__["__parse_types"] is original_parse_types
+    assert Generator.UNWRAPPED_INTERVAL_VALUES == unwrapped_interval_values
+    assert Generator.TRANSFORMS == transforms
+    assert set(Tokenizer.VAR_SINGLE_TOKENS) == var_single_tokens
+
+    assert parse_one("SELECT CAST(1 AS INT)").sql() == "SELECT CAST(1 AS INT)"
+    assert d.parse_one("SELECT @x AS y").sql() == "SELECT @x AS y"