diff --git a/CHANGELOG.md b/CHANGELOG.md
index 70cc02a..e7f7e73 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -49,6 +49,7 @@
 - [#164](https://github.com/clojure-emacs/clojure-ts-mode/pull/164): Support `setopt-local` for `clojure-ts-indent-style`, `clojure-ts-extra-def-forms` and `clojure-ts-comment-macro-font-lock-body`, which Emacs 31 refused to set locally and Emacs 32 warned about in `.dir-locals.el`.
 - [#168](https://github.com/clojure-emacs/clojure-ts-mode/pull/168): Indent docstrings like `clojure-mode`, moving the lines indented less than `clojure-ts-docstring-fill-prefix-width` to that column and keeping deeper ones, instead of leaving them alone with `clojure-ts-use-markdown-inline` and flattening them without it.
 - [#174](https://github.com/clojure-emacs/clojure-ts-mode/pull/174): Speed up indentation of forms with many children, like a big `comment` form, which slowed down with the cube of their length.
+- [#175](https://github.com/clojure-emacs/clojure-ts-mode/pull/175): Fix `clojure-ts-align` signalling `treesit-node-outdated` on a line that ends in a form after metadata and more than one space.
 
 ## 0.6.0 (2025-12-02)
 
diff --git a/clojure-ts-mode.el b/clojure-ts-mode.el
index 17049c7..d919ddb 100644
--- a/clojure-ts-mode.el
+++ b/clojure-ts-mode.el
@@ -1896,19 +1896,21 @@ children of the rich comment form are also considered as top-level."
   "Return the point that represents the end of the current defun."
   (treesit-node-end (treesit-defun-at-point)))
 
-(defun clojure-ts--search-whitespace-after-next-sexp (root-node bound)
+(defun clojure-ts--search-whitespace-after-next-sexp (root-node-fn bound)
   "Move the point after all whitespace following the next s-expression.
 
 Set match data group 1 to this region of whitespace and return the
 point.
 
 To move over the next s-expression, fetch the next node after the
-current cursor position that is a direct child of ROOT-NODE and navigate
-to its end.  The most complex aspect here is handling nodes with
-metadata.  Some forms are represented in the syntax tree as a single
-s-expression (for example, ^long my-var or ^String (str \"Hello\"
-\"world\")), while other forms are two separate s-expressions (for
-example, ^long 123 or ^String \"Hello\").  Expressions with two nodes
+current cursor position that is a direct child of the node ROOT-NODE-FN
+returns and navigate to its end.  It's called again for every line, as
+fixing the space after metadata on the way leaves the old node outdated.
+The most complex aspect here is handling nodes with metadata.  Some
+forms are represented in the syntax tree as a single s-expression (for
+example, ^long my-var or ^String (str \"Hello\" \"world\")), while
+other forms are two separate s-expressions (for example, ^long 123 or
+^String \"Hello\").  Expressions with two nodes
 share some common features:
 
 - The top-level node type is usually sym_lit
@@ -1927,7 +1929,8 @@ BOUND bounds the whitespace search."
             (progn
               (search-forward-regexp regex bound)
               (point))
-          (when-let* ((cur-sexp (treesit-node-first-child-for-pos root-node (point) t)))
+          (when-let* ((cur-sexp (treesit-node-first-child-for-pos
+                                 (funcall root-node-fn) (point) t)))
             (goto-char (treesit-node-start cur-sexp))
             (if (clojure-ts--metadata-node-p cur-sexp)
                 (progn
@@ -1943,7 +1946,7 @@ BOUND bounds the whitespace search."
               ;; End-of-line, try again on next line.
               (`(12) (progn
                        (forward-char 1)
-                       (clojure-ts--search-whitespace-after-next-sexp root-node bound)))
+                       (clojure-ts--search-whitespace-after-next-sexp root-node-fn bound)))
               ;; Closing paren, stop here.
               (`(5 . ,_) nil)
               ;; Anything else is something to align.
@@ -2112,8 +2115,10 @@ between BEG and END."
           (clojure-ts--point-to-align-position sexp-type node)
           (align-region (point) node-end nil
                         `((clojure-align (regexp . ,(lambda (&optional bound _noerror)
-                                                      (let ((updated-node (clojure-ts--node-from-sexp-data beg (marker-position end) sexp)))
-                                                        (clojure-ts--search-whitespace-after-next-sexp updated-node bound))))
+                                                      (clojure-ts--search-whitespace-after-next-sexp
+                                                       (lambda ()
+                                                         (clojure-ts--node-from-sexp-data beg (marker-position end) sexp))
+                                                       bound)))
                                          (group . 1)
                                          (separate . ,clojure-ts-align-separator)
                                          (repeat . t)))
diff --git a/test/clojure-ts-mode-indentation-test.el b/test/clojure-ts-mode-indentation-test.el
index 796fa75..c12e854 100644
--- a/test/clojure-ts-mode-indentation-test.el
+++ b/test/clojure-ts-mode-indentation-test.el
@@ -1482,4 +1482,12 @@ c)
     :foo   \"bar\"}"
 
     "#::alias{:a   1
-         :bbb 2}"))
+         :bbb 2}")
+
+  (it "should align a line ending in a form after metadata and extra spaces"
+    ;; Fixing the space after the metadata used to leave the node it
+    ;; searched outdated for the next line.
+    (let ((clojure-ts-align-forms-automatically t))
+      (with-clojure-ts-buffer "{:a ^:m  b\n:ccc d}"
+        (indent-region (point-min) (point-max))
+        (expect (buffer-string) :to-equal "{:a   ^:m b\n :ccc d}")))))