It's prudent to just never buy stuff like that on ebay. It may be ok for tinkering around on a breadboard but you can't trust the specs and really, why not just buy from a reputable seller like Digikey? They are still fairly inexpensive and much more likely to be real genuine parts that meed the published specs.
You're right, but if only it was that easy (to always be able to buy from digikey or equivalents).
What I do is hobby related and sometimes I do small runs of 5~10 units for sale too so this affects how I buy parts.
Here(in my part of EU) we have 3 large distributors Digikey, Mouser and Farnell. Out of those 3 only Farnell has a local warehouse. Shipping is still expensive, but cheaper than the remaining two. Other than those 3 we also have lots of little family run shops that often have been in business for decades. They are usually quite reputable. Finally there are various auction and local ads sites (like ebay/Craigslist).
The cost of shipping for those little shops is usually very cheap, double that for Farnell, double that again for Mouser, and double it for the third time for Digikey. Most of my order values are around what it costs to ship with digikey, but that is not the only issue with large distributors. These days it is very difficult to get all parts at one distributor. So imagine having to order with all 3.
Also, the kind of older parts I often buy (74LS,STM32,Atmega etc) are sometimes found on local ad sites sold by individuals.
So long story short, if you're making commercial projects and you have multi $k budget it is indeed prudent to only buy from big distributors. For small scale buyers... We sometimes have to take chances.
Here is an update about those counterfeit chips as I've spent some time on a breadboard testing them and comparing with a genuine thing (bought in a local fairly reputable shop). Interestingly they can be made to work even at rather high speeds, but they have a much stronger tendency to self oscillate and the propagation delay is about 3 times the genuine thing.
I used a breadboard as everyone knows those things are not ideal for signal integrity, but still there are lots of different behaviours between those two chips so let's list them. C is counterfeit, G is genuine
1.
C - Applying power to both chips with all inputs and outputs pulled down with 12k to ground (no decoupling cap) makes the chip oscillate immediately on two out of its 4 gates. The noise of this oscillations is present even on outputs of two gates that don't oscillate (at 1V).
G - No oscillations
2. 100nF ceramic decoupling cap added across power right next to chip.
C - no difference
G - no difference
3. Outputs of all gates have been switched from resistor pull downs to pull up(same 12k) to be in line with nand truth table.
C - Self oscillations gone
G - no difference
4. Propagation delay measured.
C - Approx 38nS
G -Approx 12nS (still worse than datasheets value of 9)
5. Square ttl signal was applied to one of each gate's inputs. Frequency was varied from 10kHz to 40Mhz.
C - works fine until it starts drawing a lot more current at about 15MHz, 50mA PSU current limit cuts off power at around 23MHz
G - Works fine all the way. Power consumption slightly rises from "0" to 3mA near 20MHz,Output amplitude drops slightly above 30MHz. I can't say if this is a chips fault or the breadboard wiring (likely the latter).
6. Sine wave applied same frequency range (10k to 40MHz).
C - Same effect as with square wave above 15MHz,below however the shape of the output signal looks "better" than the genuine thing? Strange.
G - Fine all the way, but at mid frequencies of 1MHz the shape of the output looks a little strange. When it switches high it reaches well above threshold (about 4V),then slowly raises to 5V. The counterfeit chip went straight to 5V (when it worked).
So here we have it. I suspect those are not even TI chips (on account of the propagation delay being so much higher), but some other make. Perhaps they are rejects from manufacturing that failed self oscillation/crosstalk tests. Do logic manufacturers do such tests? I have no idea.