You probably aren't testing with sufficiently random noise to see problems. You certainly aren't testing to find the noise margin, making this observation nearly meaningless.
![Tongue :P](https://www.eevblog.com/forum/Smileys/default/tongue.gif)
Paranoia about bypass caps is largely overrated, anyway. The advice is repeated out of ignorance, and usually winds up doing more good than harm, so in the absence of proper analysis, it's the best minimum.
But there are absolutely situations where applying the advice actively harms the circuit. Nothing can replace properly understanding and analyzing the power supply network!
Back when I built this thingy,
![](https://www.seventransistorlabs.com/Images/Z80_Timer2.jpg)
It would run ~forever with a repetitive program, but crash randomly (within days) when I added a LFSR routine.
Additional supply bypass didn't have any effect on stability, but that is obvious from PSN analysis -- the +/- wires are always nearby, so the impedance between them is small, and the effect between regions is small (the PSN has a low impedance, so there is little ripple propagating through it). No, the problem is the PSN shape relative to the routed signals, and the routed signals themselves which are through air, high impedance (>100 ohms), and not routed in parallel with the PSN, but across rungs of it.
Most likely failure mode would be: a bunch of (data or address) lines all change at once, causing a large enough displacement current along the bus, in turn causing signal bounce between the local supply (which is "cantilevered" out on a rung, here) and the bus signals. It doesn't take much bounce to violate the TTL V_IL (0.8V).
In effect, you get a series resonant circuit, between the supply inductance (of the +/- leads in parallel), input pin capacitances (all acting in parallel), bus wires (all acting in parallel), and all pin drivers (acting in parallel and driving the circuit).
74HC is somewhat faster than TTL, lower current consumption, and symmetrical logic threshold (= more noise immunity). It's fast enough to cause signal quality issues on modest length runs (~1m+), and questionable behavior on a breadboard. You're lucky to escape problems with a board that size -- but, you may also be just that: lucky. Without checking actual signal quality, who knows?
Tim