Regarding BJTs, google it: "The [LTSpice] bipolar junction transistor model is an adaptation of the integral charge control model of Gummel and Poon. This modified Gummel-Poon model extends the original model to include several effects at high bias levels, quasi-saturation, and substrate conductivity.
The model automatically simplifies to the Ebers-Moll model when certain parameters are not specified."
So, Ebers-Moll is the
worst it can do
.
One disparity between simulation and reality is obvious: in simulation, all wires are perfect zero-length superconductors, and parasitic capacitance does not exist. Reality is... more complex than that. A inherently stable, well-designed circuit will tend to be stable both in simulation and reality; a very unstable, poorly-designed circuit will tend to be unstable in both simulation and reality. Where it gets tricky is when your design is right on the very bleeding edge of stability -- reality and simulation will be very sensitive to initial conditions, and thus they might "disagree" from a casual observer's perspective. With simulation, you can easily do Monte Carlo analysis to try a whole bunch of component values (e.g. resistor/capacitor values +/- 10%) to check that your design is robust. None of this really answers your question, but should give a flavour of what to expect. I mean, sometimes if you build "exactly the same circuit" twice in reality, one will oscillate like crazy and the other will be fine -- what is the simulation supposed to do to tell you this?
Another reason that simulation
can be unfairly disliked is that whenever a simulation disagrees with intuition, people tend to blame the simulation. For example, in one of Dave's videos he made a circuit involving a (nominally little, low-power, low gate threshold) MOSFET and in the LTSpice simulation the MOSFET never turned on. Much hate was thrown at the simulation, but the thing is the specific transistor model was randomly chosen from a list, and a check of the relevant datasheet showed that far from a little SOT-23 thing, it was a monster power MOSFET intended to be driven with 12V gate drive. The simulation was absolutely perfectly correct, and yet was labelled as being completely wrong. So for reasons like these, I'm suspicious of people who label simulations as bad; I'll be the first to agree that they require a pedantic level of care to prevent such errors (I mean, in real life, you'd take a moment to check the gate threshold if you pulled a huge MOSFET out of your parts bin). But at the same time, reality is very complex, and no-one's claiming that a simulation will correctly simulate an oscillation arising from the lack of supply decoupling caps (since you have perfect zero-length superconductors connected to your perfect voltage source, decoupling caps just don't make sense in a simulation).
Anyway, far from a yes/no answer, but I hope that's helpful