I am skeptical too. Power usage on the 3v3 rail, when drv is under load, is around 20mA. Not sure if there was some kind of problem with the voltage regulator.
Measuring the supply voltages with a scope can show noise levels, but also if there is high frequency oscillation due to the lack of proper decoupling. I have seen this happen in test circuits without decoupling and could solve it by adding a proper decoupling cap. Usually 100nF does the trick.
Replacing the LDO with an other one could have changed the dynamics of the circuit and solved the problem that way.
Yep, you definitely want to add a decoupling cap for the microcontroller - a 100nF ceramic (MLCC, X5R/X7R, 16V) will do fine. You always want to use one for each separate power pin of the uC and put them very close to the pins. See the
AVR040 application note for more information on EMC design (read chapter 4.6, or better yet, the whole document).