Discrepancy between output balancing and input balancing of reversed graph

[alegnani]

Bug detailed here.

This 2-5 balancer is shown as being input unbalanced:

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

The same balancer but manually reversed, yielding a 5-2 balancer is not proven to be output unbalanced:

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