I conjecture two possibilities.
Incapable of rational resolution due to equality of evidence or argument for and against
As regards antinomies I would hazard 'incapable of rational resolution'. In an antinomy there are two contradictory hypotheses, theories or claims and each is confirmed by equally cogent evidence or argument. Moreover, the equal confirmation of each hypothesis, etc., is such that no conceivable extra evidence or new argument could tilt the balance in favour of one hypothesis or the other. That, I think, is the sense in which an antinomy is said or supposed to be 'undecidable'.
But it is not to say that there are any such antinomies in philosophy or science. (It's 'highpriori' to suppose that no new argument could resolve an antinomy.)
Consistent with all possible evidence
Metaphysics is not confined to antinomies, of course. Other metaphysical issues are also said by some to be undecidable. Take a claim such as that 'The concrete universal determines its own particularisation', which is what the late 19th-century Oxford Idealists such as Bernard Bosanquet said (M.B.Foster, 'The Concrete Universal: Cook Wilson and Bosanquet', Mind, Vol. 40, No. 157 (Jan., 1931), pp. 1-22: 1). This certainly looks like a metaphysical claim; and it sits within the complex metaphysical system of (Oxford) Absolute Idealism.
However, what evidence could one deploy to prove or defend it? Put another way and with acknowledgement to William James: How would the world be different if it were false? Since we can't say or even begin to answer this question, we can designate it as 'undecidable'. That is, if we do not take the different path of saying that it is meaningless.