How are you measuring VCC? If you are using a DC voltmeter it may be that VCC is dipping way lower than 4.7V then back up -- in a microprocessor system this would be a disaster.
Also, do you have adequate decoupling? There is some discussion here about how little you can get away with, but for TTL and slow MOS the standard was a single 10nF (.01uF) cap across each chip's power and ground leads, as close as you can get, plus some overall board decoupling in the form of a 10uF electrolytic. On a proto board you want to keep the leads _short_, on a PC board you want to keep things compact as well.
Leaving off your decoupling caps isn't courting disaster -- it's going to her daddy's farm and abducting her in broad daylight.