Returning to matters OPAQUE, here is the following problem which may well now be approachable by known methods.
Let me phrase the conjecture in the case when the prime p = 2 and the level N = 1.
As we know from Buzzard-Kilford, in every classical weight \kappa “close enough” to the boundary of weight space, the slopes of the space of overconvergent forms are given by the arithmetic progression nt where t depends only on the 2-adic valuation of \kappa(5) – 1. Now, for each of these overconvergent forms, one obtains a Galois representation
\rho_{n}: G_{\mathbf{Q}} \rightarrow \mathrm{GL}_2(\overline{\mathbf{Q}}_2)
for every positive integer n. This gives a map from the integers \mathbf{N} considered as a discrete set to \mathrm{Spf}(R) for a deformation ring R (there is only one residual representation in this setting. Yes, it is residually reducible, but ignore this for the moment).
Problem: Show that this map from \mathbf{N} extends to a continuous map from \mathbf{Z}_2.
I’ve never done any computations in these weights, but my spidey senses says it should be true. Naturally, one should also try to work out the most precise statement where N and p are now arbitrary.
I don’t have any sense about is whether, for a fixed weight \kappa, there is actually a representation
\rho: G_{\mathbf{Q}} \rightarrow \mathrm{GL}_2(\mathcal{O} [[T]])
(for some \mathcal{O} containing enough roots of unity) whose specialization to T = n for a non-negative integer n is \rho_n, or whether the continuity is not so strong. That might be interesting to check.
More natural questions:
1. Once one has the correct formulation in fixed weight \kappa, explain what happens over the entire boundary, and at the halo.
2. I’m pretty sure that \rho_0 will just be the Eisenstein series, or more accurately the Galois representation 1 \oplus \chi, where \chi is determined from \kappa in some easy way involving normalizations which I don’t want to get wrong. But what is \rho_{-1}? I’m not sure if it is interesting or not. But is there any way of parametrizing this family of Galois representations so that the potentially crystalline points transparently correspond to non-negative integers?
All of this is just to say that, even for N = 1 and p = 2, there’s a lot we don’t know about the eigencurve over the boundary of weight space.
I would like to point to a recent paper (Section 3.3)
http://people.maths.ox.ac.uk/vonk/documents/p_tadic.pdf
where the author observes that his computations suggest that a similar continuity holds for boundary forms.
Excellent — Jan was in my AWS group which touched on these questions. It certainly confirms numerically everything one would hope for. But now either an explanation or a proof would be nice!