diff --git a/src/parser/gjs-gts-parser.js b/src/parser/gjs-gts-parser.js
index d8213a1..ab70f90 100644
--- a/src/parser/gjs-gts-parser.js
+++ b/src/parser/gjs-gts-parser.js
@@ -232,7 +232,7 @@ export function parseForESLint(code, options) {
             range: c.range,
             loc: c.loc,
           })),
-        ];
+        ].sort((a, b) => a.range[0] - b.range[0]);
       }
     }
 
diff --git a/tests/program-comments-sort-order.test.js b/tests/program-comments-sort-order.test.js
new file mode 100644
index 0000000..1a5e63d
--- /dev/null
+++ b/tests/program-comments-sort-order.test.js
@@ -0,0 +1,87 @@
+/**
+ * Regression test for the sorted-by-range invariant on Program.comments.
+ *
+ * ESLint's SourceCode builds `tokensAndComments = sortedMerge(tokens, comments)`
+ * and `createIndexMap(tokens, comments)` — both iterate comments and tokens
+ * with a merge that assumes each array is already sorted by `range[0]`. Every
+ * standard JS parser (espree, @babel/eslint-parser, @typescript-eslint/parser)
+ * honors that invariant.
+ *
+ * When a .gts file has a JS block comment (slash-star style) interleaved between templates,
+ * TS-parser comments are spread into `program.comments` first and Glimmer
+ * template comments get appended — producing an array like
+ * `[jsAt56, glimmerAt23]` whose order doesn't match range order. The
+ * downstream effect is `sourceCode.getTokenBefore(glimmerComment)` /
+ * `getTokenAfter(glimmerComment)` returning wrong tokens (or tokens from
+ * entirely different template regions), because the `indexMap` keyed on
+ * unsorted input points at the wrong token index.
+ */
+
+import { describe, expect, it } from 'vitest';
+import { Linter } from 'eslint';
+import { parseForESLint } from '../src/parser/gjs-gts-parser.js';
+
+describe('program.comments sort order (ESLint tokensAndComments invariant)', () => {
+  const mixedSource = [
+    'const X = <template>',
+    '  {{! glimmer comment at 22 }}',
+    '</template>;',
+    '/* js comment at 56 */',
+    'const Y = 1;',
+  ].join('\n');
+
+  it('ast.comments is sorted by range[0]', () => {
+    const { ast } = parseForESLint(mixedSource, {
+      filePath: 't.gts',
+      range: true,
+      loc: true,
+      comment: true,
+      tokens: true,
+    });
+    const starts = (ast.comments || []).map((c) => c.range[0]);
+    const sorted = [...starts].sort((a, b) => a - b);
+    expect(starts).toEqual(sorted);
+  });
+
+  it('getTokenBefore / getTokenAfter on a Glimmer comment return source-adjacent tokens', () => {
+    const linter = new Linter();
+    linter.defineParser('p', { parseForESLint });
+    const probes = [];
+    linter.defineRule('probe', {
+      create(context) {
+        return {
+          'Program:exit'() {
+            const sc = context.sourceCode;
+            for (const c of sc.getAllComments()) {
+              if (c.value.includes('glimmer')) {
+                probes.push({
+                  commentRange: c.range,
+                  before: sc.getTokenBefore(c)?.range ?? null,
+                  after: sc.getTokenAfter(c)?.range ?? null,
+                });
+              }
+            }
+          },
+        };
+      },
+    });
+    linter.verify(
+      mixedSource,
+      {
+        parser: 'p',
+        parserOptions: { ecmaVersion: 2022, sourceType: 'module' },
+        rules: { probe: 'error' },
+      },
+      { filename: 't.gts' }
+    );
+    expect(probes).toHaveLength(1);
+    const { commentRange, before, after } = probes[0];
+    // Whatever the exact adjacent token ranges are, they must bracket the
+    // comment — the token before must end at or before the comment's start,
+    // and the token after must start at or after the comment's end.
+    expect(before).not.toBeNull();
+    expect(after).not.toBeNull();
+    expect(before[1]).toBeLessThanOrEqual(commentRange[0]);
+    expect(after[0]).toBeGreaterThanOrEqual(commentRange[1]);
+  });
+});