Can you provide your model as test case to reproduce the issue?

Unfortunately, I cannot provide the model as is. It is an entire package and I'm not allowed to share the code anyway. But I would want to find a way to debug it, so that I can generate a minimal failing model that I can send to you. Could you guide me through some debugging techniques?

The OpenModelica commit pasted above is actually incorrect, here is the correct one: 7dfbb017b3da1f05671a15989aff0799ffdf2632

Well, I did quite a lot of changes to the backend in that specific commit. Therefore, it is hard to pinpoint something.
However, we could start with the error message ;-). You wrote that your model failed during simulation, right? Is there any error message or something? Did you run the simulation from command line or using OMEdit?

Yes, it simulates and then breaks with this error message:

stdout            | warning | The default linear solver fails, the fallback solver with total pivoting is started at time 0.000000. That might raise performance issues, for more information use -lv LOG_LS.
DASKR--  AT T (=R1) AND STEPSIZE H (=R2) THE                                   
      In above,  R1 =   1.4663589072380E+00   R2 =   3.7410878546409E-16
DASKR--  NONLINEAR SOLVER FAILED TO CONVERGE                                   
DASKR--  REPEATEDLY OR WITH ABS(H)=HMIN                                         
stdout            | warning | The corrector could not converge.
stdout            | warning | can't continue. time = 1.466359
assert            | warning | The following assertion has been violated at time 1.466359

<Here some variables go out of their min max ranges>

stdout            | info    | model terminate | Integrator failed. | Simulation terminated at time 1.46636

Running fro the command line with omc and these notable options:

cflags="-O3",
tolerance=1e-6

With LOG_LS I get this (amongst others):

LOG_LS            | warning | Failed to solve linear system of equations (no. 3045) at time 1.466359, system is singular for U[2, 2].
LOG_LS            | info    | Matrix U
|                 | |       | |          0
LOG_LS            | info    | Output vector x
|                 | |       | | [ 0]                    0
LOG_LS            | warning | The default linear solver fails, the fallback solver with total pivoting is started at time 1.466359. That might raise performance issues, for more information use -lv LOG_LS.
LOG_LS            | info    | Start solving Linear System 3045 (size 1) at time 1.46636 with Total Pivot Solver
LOG_LS            | info    | ###  0.000000  time to set Matrix A and vector b.
LOG_LS            | warning | Matrix singular!
LOG_LS            | info    | rank =  0

Any ideas? It looks like OM thinks the system is overdetermined.

Sorry, I cannot give you any hint. Can you try to isolate the failing linear system and set up a small model that can be used as public test case?

Thanks. I will try my best. It may take some time.

I've filed an issue with attached failing models:

https://trac.openmodelica.org/OpenModelica/ticket/3628