Mutate() issue - Hierarchical Forecasting

Hi, I am reposting this issue on GitHub, with a more complete example, as I suspect it might not be related to the data being used or code mistakes.

I am trying to perform Hierarchical Forecasting on a dataset that is fundamentally structured in the same way as the tourism tsibble referenced in Forecasting: Principles and Practice, but with more hierarchical levels. However, after the structural aggregation, a mutate() error shows up.
The data doesn't contain any missing values.

Following, you will find a reprex of the code, containing a minimal version of the data used that is able to reproduce the error.

Thanks in advance.

library(fable)
library(dplyr)
library(tsibble)
library(tidyverse)

t_london <- tibble::tribble(
  ~Month,             ~Value.type,   ~LSOA11CD,             ~LSOA11NM,     ~WD19CD,      ~WD19NM,    ~LAD19CD,         ~LAD19NM,           ~CTYNM, ~RGN19NM, ~CNTY21NM, ~NTN21NM, ~Count,
  "2010 Dec", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     2L,
  "2011 Jan", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     2L,
  "2011 Feb", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     3L,
  "2011 Mar", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     2L,
  "2011 Apr", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     0L,
  "2011 May", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     2L,
  "2011 Jun", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     4L,
  "2011 Jul", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     3L,
  "2011 Aug", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     2L,
  "2011 Sep", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     0L,
  "2011 Oct", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     1L,
  "2011 Nov", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     1L,
  "2011 Dec", "Value-Type-1         ", "E01000001", "City of London 001A", "E05009288", "Aldersgate", "E09000001", "City of London", "City Of London", "London", "England",     "UK",     6L
)

t_london <- t_london  %>%
mutate(Month = yearmonth(Month)) %>%
  as_tsibble(key = c(LSOA11CD, Value.type), index=Month)

london_full <- t_london %>% aggregate_key((NTN21NM/ CNTY21NM / RGN19NM / CTYNM / LAD19NM / WD19NM /LSOA11NM) * Value.type, Total = sum(Count))

fit <- london_full %>%
  model(base = ARIMA(Total)) %>%
  reconcile(
    bu = bottom_up(base),
    ols = min_trace(base, method = "ols"),
    mint = min_trace(base, method = "mint_shrink"),
  )
#> Warning in max(which(abs(ma) > 1e-08)): no non-missing arguments to max;
#> returning -Inf

#> Warning: 16 errors (1 unique) encountered for base
#> [16] argument must be coercible to non-negative integer

fc <- fit %>%
  forecast(h = 5)
#> Warning: Problem with `mutate()` input `mint`.
#> ℹ diag(.) had 0 or NA entries; non-finite result is doubtful
#> ℹ Input `mint` is `(function (object, ...) ...`.
#> Warning: Problem with `mutate()` input `mint`.
#> ℹ diag(.) had 0 or NA entries; non-finite result is doubtful
#> ℹ Input `mint` is `(function (object, ...) ...`.
#> Error: Problem with `mutate()` input `mint`.
#> x infinite or missing values in 'x'
#> ℹ Input `mint` is `(function (object, ...) ...`.

^{Created on 2021-02-10 by the reprex package (v0.3.0)}

Start with the supplied R reprex, especially aggregate_key(), reconcile(), and forecast(h = 5), and reproduce the mutate() failure and preceding ARIMA warnings. Trace the mint reconciliation path and add a regression test for this hierarchical structure; done means forecasting completes without the reported non-finite-value error.