This is a synchronic answer, pertaining only to classical Latin, as I believe the use of gerunds and gerundives was subject to some change praeclassically; what is more, they may have once been the same thing, or one may have been born from the other. So it would not make sense for me, especially without enough information, to distinguish between them in archaeic Latin or earlier. Looking at agitandum est vigilias, one might call that a mixed Ur-form of sorts.
Summary of my position: this construction is generally considered by classicists to be a nominative gerundive.
The main reason is its modality: there is no way to translate nunc est bibendum without a flavour of obligation or similar; gerunds by themselves never have a sense of obligation. This suggests that it cannot be a gerund. As a check, I went through all the instances of bibendum in the HP corpus where it is used in the relevant construction: in each case, a modal translation seems to be required. Example:
uitandi acres cibi, lacte uinoque pingui utendum, bibendumque liberalius quam edendum est
"one must drink more liberally than one must eat" (not: "one drinks more liberally")
— Celsus, De Medicina 6.6.15b
Your translation "?now it is time to drink" does not seem possible to me: how can a translation "time to" be found within the gerund? One would rather translate it as "*this is drinking" or "*drinking is thus".
In addition, an -ndum word is never used as the subject of a verb with a non-modal translation: to express "drinking is this", I believe classical authors always used the infinitive, bibere hoc est. (The same applies to direct objects, which are never gerunds but always infinitives, i.e. incipio bibendum → incipio bibere.) Example:
radicem decoctam bibere spasticis, eversis, convulsis, suspiriosis salutare est
"drinking root extract is salutary" (notice the agreement of salutare with the neuter infinitive)
— Plinius the Elder, Naturalis Historia 21.132
In other words: whenever you come across an -ndum form as the subject of a verb, I believe you will find that you need a modal translation of sorts. In yet other words: the only time writers used the construction in question is precisely when they wanted to express this modality; when they didn't, they used the infinitive.
As to passivity, it is not uncommon to use the passive in impersonal constructions, with a neuter participle where applicable: Romam itur "one goes to Rome" (or, more liberally, "we/they/etc. go to Rome"). Cognitum est "it was thought". This is the same type of construction, for we have 1.) a passive, 2.) a neuter nominative, and 3.) no agent except as a dativus auctoris (cf. [uncommon] cognitum est nobis "we thought"). It might even be a feature of Indo-European, cf. Dutch er werd gegaan "[there was gone]" => "one went", and English "it was decided that...".
Its similarity to (other) gerundive constructions also points in the gerundive direction:
bibendum est "[there is to be drunk]" (= "one must drink")
bibendum est vinum "wine is to be drunk" (= "one must drink wine")
bibendum vinum dedi "I gave wine to be drunk" (= "I gave wine that must be drunk / to drink")
All of these are common, and they share the same praedicate frames except where complements are absent. Any substantive noun agreeing with the -ndum word takes the same thematic role as would the object in an active sentence: the role is theme here, vinum bibit → vinum bibendum est. This is itself a clear sign of passivity (cf. passive participles). Any dative takes the same thematic role as would the subject in an active sentence: the role is agent here, nos vinum bibimus → nobis vinum bibendum est). The dativus auctoris is another sign of passivity. When you apply the construction to some other verb, you will notice the same telling transformation of praedicate frames. This is not at all similar to (other) gerund constructions, which have very different transformations of the praedicate frame.
As to its case, the nominative would seem our only option, since the verb sum takes only nominative complements. Further, when you keep the same praedicate frame but add a subject, such as vinum or pocula, the -nd- form changes along with it, always to agree with the subject, in the nominative: pocula bibenda sunt.
It is probably best considered a dominant construction (à la Caesar occisus senatoribus placet); for the real message here is not "there is wine, which must be drunk", but "the drinking of wine must be done". This is just as in other gerundive constructions. It also explains perceived similarities to the gerund: mordebatur in floribus carpendis ("she was bitten during the picking of flowers", not "during the flowers to be picked") means the same thing as in flores carpendo because the gerundive, an adjective, is used dominantly there, i.e. as if it were a substantive noun—and a substantive "plucking" is a gerund. Dominancy is a bit more difficult to sense in the impersonal construction; but, if we did assume this construction to be dominant as well, that would make the construction seem more 'logical'.
As to translations, one often supplies a logical agent from context: "Now let us drink!". An imperative is also possible: "Now drink!". Add "all ye" for fun?