Roland Steiner wrote:

Cf. Böhtlingk/Roth (PW) s.v. nagara: "Das na wird in keinem comp. ṇa
nach gaṇa kṣubhnādi zu P. 8, 4, 39." ("The na is not cerebralized in a
compound according to gaṇa kṣubhnādi ...").

Example: giri-nagara "mountain-city" (in contrast to
giri-ṇadī/giri-nadī "mountain-torrent").

Why then do we have sūrya-nadī instead of sūrya-ṇadī,(from Monier-Williams) ?

Similarily also from Monier-Williams: why sūrya-nakṣatra, sūrya-nārāyaṇa, sūrya-namaskāra, sūrya-netra instead of , sūrya-ṇakṣatra, sūrya-ṇārāyaṇa, sūrya-ṇamaskāra, sūrya-ṇetra

I've just now seen what Whitney writes: "This rule applies strictly and especially when the nasal and the cause of its alteration lie within the limits of the same integral word; but it is extended , within certain limits, to compound words'.

What does he mean by "[applies] within certain limits to compound words." Does he mean all compounds not containing a word from gaṇa kṣubhnādi or are there other restrictions to its application in compounds?

Thanks,

Harry Spier