VVV-WIT-01: highly obscured classical nova or protostellar collision?

A search of the first Data Release of the VISTA Variables in the Via Lactea (VVV) Survey discovered the exceptionally red transient VVV-WIT-01 ($H-K_s=5.2$). It peaked before March 2010, then faded by $\sim$9.5 mag over the following two years. The 1.6--22 $\mu$m spectral energy distribution in March 2010 was well fit by a highly obscured black body with $T \sim 1000$ K and $A_{K_s} \sim 6.6$ mag. The source is projected against the Infrared Dark Cloud (IRDC) SDC G331.062$-$0.294. The chance projection probability is small for any single event ($p \approx 0.01$ to 0.02) which suggests a physical association, e.g. a collision between low mass protostars. However, black body emission at $T \sim 1000$ K is common in classical novae (especially CO novae) at the infrared peak in the light curve, due to condensation of dust $\sim$30--60 days after the explosion. Radio follow up with the Australia Telescope Compact Array (ATCA) detected a fading continuum source with properties consistent with a classical nova but probably inconsistent with colliding protostars. Considering all VVV transients that could have been projected against a catalogued IRDC raises the probability of a chance association to $p=0.13$ to 0.24. After weighing several options, it appears likely that VVV-WIT-01 was a classical nova event located behind an IRDC.