That behaviour is explicitly defined here:

<ST_IN_SCRIPTING>"#"|"//" {
        while (YYCURSOR < YYLIMIT) {
                switch (*YYCURSOR++) {
                        case '\r':
                                if (*YYCURSOR == '\n') {
                                        YYCURSOR++;
                                }
                                /* fall through */
                        case '\n':
                                CG(zend_lineno)++;
                                break;
                        case '?':
                                if (*YYCURSOR == '>') {
                                        YYCURSOR--;
                                        break;
                                }
                                /* fall through */
                        default:
                                continue;
                }

                break;
        }

        yyleng = YYCURSOR - SCNG(yy_text);

        RETURN_TOKEN(T_COMMENT);
}

Removing the "case '?':" clause would break the following inline code:

<?php echo "Hello, World!"; // my comment ?> Remainder

Which should print: Hello, World! Remainder

Everything on the same line after the comment including after the closing tag are commented out.

I understand this request.  PHP ought to know if this is a single-line execute, and handle accordingly.

  <?php echo "hello"; // comment ?> output
and
  <?= echo "hello"; // comment ?> output
should each run with output:
  hello output

That said, the request here:
<?php
  if (foo) {
    // ?>
  }
}
?>

Should not throw a syntax error.  I have no idea what the consensus on this kind of issue is, however, (i've done) and it would be simple enough to track what line <?php, or <?, or <?= start on, and what line we see // ?>, and if they aren't the same, then ignore the ?> inside the comment.

ie:
if (*YYCURSOR == '>' && (CG(zend_opentag_lineno) == CG(zend_lineno))) {

Since this is against a security fix only branch, and since a patch against a supported branch would need to be different, please can you open a PR on github against a supported branch.