nicw:
I measured 49mA on one board, and 44mA on one other
Even on your boards with the issue, both are pushing current above the 40ma maximum DC Current per I/O Pin
as specified in the AVR absolute maximum ratings table.
So I have a very hard time understanding how anyone, especially the goofy designers of
the board that you contacted, could disagree that this is a problem, particularly if you read
the AVR documentation which explicitly states:
Stresses beyond those listed under “Absolute Maximum Ratings”
may cause permanent damage to the device.
But regardless of whether you are able to convince one vendor to correct his broken design,
it doesn't fully resolve the problem going forward, because there are several vendors
out there making similar LCD keypad shields and many of them appear to have copied
the same broken backlight circuit design so the problem will continue to live on.
--- bill