“Haber” agrees with its object
There are two dialects with respect to agreement in Spanish Existential Constructions (SEC) with haber (have, be): SI and SII. In SI the verb has default agreement (3er person, singular):
(1) Había dos hombres en la fiesta SI
but in SII, the verb agrees with the internal nominal:
(2) Habían dos hombres en la fiesta SII
despite the fact that the nominal is Accusative (acc), that is, that is not the subject but the object, as shown by the cliticization ((4) from (2)). The nominal is also acc in SI, but it does not agree with the verb ((3) from (1)):
(3) Los había SI
(4) Los habían SII
This difference has been attested by several researchers, and it seems to be present in Spanish at least from the XIV century. In the present times, SI is predominant in Peninsular Spanish, whereas SII is predominant in Latin American Spanish—see for instance, Bello (1847:§781) who denounces this agreement as an “almost universal vice”. But SII is also present in Peninsular Spanish as a non standard variety, and SI is usually imposed as a prescriptive rule in Latin American Spanish—so, in Latin American Spanish, it is not hard to find educated speakers with both SI and SII.
In SI and SII, the form for present tense is normally invariable: hay [áj] , which is exceptional because the element -y (a morphological fossil from an old locative clitic) blocks the agreement. In some dialects of SII, however, the exceptional form hay [áj] becomes haen [áen] or hayn [ájan], that is, it allows the plural -n to be suffixed to the verb, unblocking the effect of -y, as reported by Kany (1951: 257) for rural Argentinean, Lapesa (1980: § 133) for Substandard Venezuelan, and Montes (1982: 384) for Colombian Antioqueño.
I have written a paper on this issue. You can read it here. Comments welcome.