Different values for MASE in between accuracy and modeltime_accuracy

Hi all, I'm not sure if there is a bug or, but think it is worth sharing that I am spotting differences in the MASE results comparing the output of accuracy() and modeltime_accuracy()

A side question I have is if Theil U and R2 make sense for appraising the accuracy of an ETS model. Usually I just look into RMSE and MASE. Many thanks.

library(tidymodels)
library(tidyverse)
library(lubridate)
library(timetk)
library(modeltime)
library(forecast)

# Data
m750 <- m4_monthly %>% filter(id == "M750") %>% select(-id)

# Split Data 80/20
splits <- initial_time_split(m750, prop = 0.8)
# <Analysis/Assess/Total>
# <244/62/306>
train_tbl <- training(splits)
test_tbl  <- testing(splits)

# --- MODELS ---
# Model 1: exp ----
model_fit_ets <- exp_smoothing(
  error  = "auto",
  trend  = "auto",
  season = "auto",
  damping = "auto"
) %>% 
  set_engine("ets") %>% 
  fit(value ~ date, data = training(splits))

# ---- MODELTIME TABLE ----

exp_models <- modeltime_table(
  model_fit_ets
)

# ---- ACCURACY ----

# * Calibration ----
calibration_tbl <- exp_models %>% 
  modeltime_calibrate(test_tbl)

accuracy_tbl <- calibration_tbl %>% modeltime_accuracy()

calibration_tbl %>% 
  modeltime_forecast(
    new_data = test_tbl,
    actual_data = m750
  ) %>% 
  plot_modeltime_forecast()

# ----

# ETS exponential smoothing state space models, from forecast
train_2 <- m750 %>% head(244)
test_2 <- m750 %>% tail(62)

train_2 <- train_2 %>% 
 select(value) %>%
  data.frame() %>%
  ts(start = c(1990, 1), frequency = 12)

test_2 <- test_2 %>% 
  select(value) %>%
  data.frame() %>%
  ts(start = c(2010, 5), frequency = 12)

ets_model<- ets(train_2)
f_ets_model <- fabletools::forecast(ets_model, new_data = test_2, h = 62)

f_ets_model
autoplot(f_ets_model)
fabletools::accuracy(f_ets_model, test_2)

# ME     RMSE      MAE       MPE     MAPE      MASE       ACF1 Theil's U
# Training set   -9.705894 154.1525 108.8509 -0.136685 1.288250 0.3475082 0.09251034        NA
# Test set     -784.869458 834.2611 786.3949 -7.621448 7.635691 2.5105769 0.53315224  1.564927  1.564927

accuracy_tbl

# A tibble: 1 x 9
# .model_id .model_desc .type   mae  mape  mase smape  rmse   rsq
# <int> <chr>       <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#   1         1 ETS(M,A,A)  Test   786.  7.64  2.34  7.32  834. 0.799

Reproduce the supplied R example using accuracy() and modeltime_accuracy(), then inspect the entry points that calculate MASE and compare their inputs and definitions. Done means the differing results are reconciled or clearly documented, with the ETS metric question addressed if it affects the reported behavior.