My understanding, is that there are two different ways, a 1N4148 could be designed into a circuit. I'm showing two extremes.
One is, that it is basically what some/Dave(EEVblog) likes to call Jelly Bean parts, such as the 1N4148. E.g. A 1N4148, being pulled to the 5V rail, via a 1K resistor, with almost 5mA flowing through it. I.e. The circuit is nowhere near any datasheet limits, and a simple jelly bean part, fits the bill just fine. That is where you could use your cheapest, worst manufacturer, riskiest supplier(s) for. Depending on how you value the quality of your final product(s).
Alternatively, it is a critical part of a circuit, really accurately, dependably and reliably needs to respond within 2.000 nanoseconds, for every pulse. In a circuit which really needs the reverse recovery time, diode capacitance and other circuit parameters, to be within certain specified limits. In which case, I'd suggest considering a specific, very high speed diode, characterized for such uses, with suitable design margin, for best quality/dependability and reliability.
Which, if it is decided to stick with a 1N4148, should then be a specific manufacturers's part (number), with a suitably good quality pedigree, behind that manufacturer.
It is exciting, using powerful test equipment, to measure tricky things. Either because the test equipment originally cost $20,000, 15 years ago, and it allows amazingly accurate and detailed tests to be performed. Or to design your own test equipment, that measures complicated parameters, with your own, home brew modifications.
But, to use either of the above equipment, to measure, what probably arguably is a generic jelly bean part, in some fairly rare, partly obscure way. Then complain, because (what seems to be), the cheapest possible supplier, from the cheapest possible component of that type they sell, then disregarding their claimed datasheet values. Then making a panic, panic, panic, a jelly bean part differs from a specification, because they measured it at 1MHz/1.5V, and gave a typical value (on the datasheet), of 4 or 5pF, rather than at 1MHz/0V, and had a maximum value of 4pF, is perhaps a little bit of an over-reaction. Sorry, if this causes offense.
As I see it, the critical part, is that it DID agree with their published datasheet, if I've followed this thread, correctly. Therefore is arguably, at least a little bit unfair, to create a complaint thread, without clearly stating that fact, right from the start.
My basic understanding, is that jelly bean parts, are NOT 100% bolted down, in overall specifications, across ALL manufacturers and available parts. They are what was easily and very cheaply being made available, at that time. E.g. 741 op-amp or 555 timer. Yes, they do have somewhat basic specifications, bolted down. But if you get into the fine detail and/or need certain parts of its specification, to be EXACTLY what you desire. You need to choose an exact/specific 741 (partly obsolete these days) or 555 timer, from a specific manufacturer. Then carefully go over their datasheet and perhaps discuss the situation with the supplying company, if your purchase volumes and/or contacts in the industry, permit.
TL;DR
No datasheet violation = Arguably no big problem, as the (specific/actual) datasheet should really be consulted, if certain parts of the component specification, really matter, in the actual use case.
There are standards, but my understanding is, that those should NOT be relied upon, too much. I.e. Datasheets should be used and/or consulting with the actual supplier(s) and their application engineers.