Quick fix 1002 suggests use of tagged switch statement inappropriately

Consider the following code, which I admit could use strings.Cut instead:

// Variable "c" is of type string.
// We wish to split it on an embedded space character, if any.
switch i := strings.IndexByte(c, ' '); {
case i == -1:
    return c, "", nil
case i == 0:
    return "", "", errors.New("value starts with a space character")
default:
    return c[:i], c[i+1:], nil
}

My LSP implementation—gopls—reports the following:

could use tagged switch on i [default]

It then offers a quick fix: "Replace with tagged switch". If I accept that suggestion, though, the resulting code is invalid, since I lose access to the "i" variable used in the return statement in the default case.

// Note the extra "i" here.
switch i i := strings.IndexByte(c, ' '); {
case -1:
    return c, "", nil
case 0:
    return "", "", errors.New("value starts with a space character")
default:
    // We can no longer use "i" here.
    return c[:i], c[i+1:], nil
}

Note also the strange insertion of an extra "i" in the initial simple statement.

It seems that there are three things wrong here:

The recommendation is unjustified, as we can't eliminate the "i" variable, given its use in the return statement.
The applied fix changes the first two case clauses, but leaves the default case clause with dangling references to a nonexistent variable.
The applied fix mangles the initial simple statement.

I don't know whether any of these are to blame on Emacs's lsp-mode that's making use of gopls.

dominikh / go-tools

Quick fix 1002 suggests use of tagged switch statement inappropriately #1613