Comment # 30 on bug 1179461 from Dr. Werner Fink

(In reply to Mischa Salle from comment #29)
> Created attachment 844356 [details]
> Fixed version of nofreeze-c72f740a.patch
> 
> Hereby a fixed version of the nofreeze-c72f740a.patch
> There are two additional return statements in _parse_part() that did not
> decrease the recurse_level leading ultimately to a too high level.
> This would happen for various attachment types such as TYPEIMAGE and
> TYPETEXT.

The diff from the patches show

-@@ -506,16 +521,16 @@ void mutt_parse_part (FILE *fp, BODY *b)
+@@ -506,15 +521,17 @@ void mutt_parse_part (FILE *fp, BODY *b)
        {
        fseeko (fp, b->offset, SEEK_SET);
        if (mutt_is_message_type(b->type, b->subtype))
@@ -117,27 +117,25 @@ reaching the limit.
        else if (ascii_strcasecmp (b->subtype, "external-body") == 0)
          b->parts = mutt_read_mime_header (fp, 0);
        else
--        return;
-+        goto bail;
++        recurse_level--;
+         return;
        }
        break;

      default:
--      return;
-+      goto bail;
++      recurse_level--;
+       return;
    }

-   /* try to recover from parsing error */
-@@ -524,6 +539,8 @@ void mutt_parse_part (FILE *fp, BODY *b)
+@@ -524,6 +541,7 @@ void mutt_parse_part (FILE *fp, BODY *b)
      b->type = TYPETEXT;
      mutt_str_replace (&b->subtype, "plain");
    }
-+bail:
 +  recurse_level--;
  }

... IMHO the current patch on obs should do the same work