BUG: `scipy.optimize.milp` sometimes crashes with `HiGHS` assertion failure in `1.9.2`

Describe your issue.

Background: just upgraded from 1.9.1 to 1.9.2 to test out the fix for #16545.

Some MILP models that solved perfectly fine with scipy.optimize.milp in 1.9.1 crash hard in 1.9.2 from deep within the HiGHS code:

Running HiGHS 1.2.0 [date: 2021-07-09, git hash: n/a]
Copyright (c) 2022 ERGO-Code under MIT licence terms

Presolve is switched off

Solving MIP model with:
   358 rows
   332 cols (101 binary, 0 integer, 0 implied int., 231 continuous)
   961 nonzeros

        Nodes      |    B&B Tree     |            Objective Bounds              |  Dynamic Constraints |       Work      
     Proc. InQueue |  Leaves   Expl. | BestBound       BestSol              Gap |   Cuts   InLp Confl. | LpIters     Time

         0       0         0   0.00%   0               inf                  inf        0      0      0         0     0.0s
         0       0         0   0.00%   111.5580767     inf                  inf        0      0      7        46     0.0s
Assertion failed: (fracval < localdom.col_upper_[col] - mipsolver.mipdata_->feastol), function selectBranchingCandidate, file HighsSearch.cpp, line 279

I tried to investigate this myself by looking into the HiGHS logs and the funny thing is that this assertion does not fail when I pass highs_debug_level with a value of 1, 2 or 3 (default=0) as a verbatim option to scipy.optimize.milp().

Also, I noticed that somehow the HiGHS version was downgraded from 1.2.2 to 1.2.0 between 1.9.1 and 1.9.2. Is this intentional? If not, I guess that might be the underlying culprit?

Reproducing Code Example

N/A

Error message

Assertion failed: (fracval < localdom.col_upper_[col] - mipsolver.mipdata_->feastol), function selectBranchingCandidate, file HighsSearch.cpp, line 279

SciPy/NumPy/Python version information

1.9.2 1.23.4 sys.version_info(major=3, minor=9, micro=13, releaselevel=‘final’, serial=0)

[mckib2]

Great! I’ll close this and it should be available in the next upcoming release

[woutervbk]

@mckib2 I can confirm that I was able to successfully run all my unit tests with https://anaconda.org/scipy-wheels-nightly/scipy/1.10.0.dev0/download/scipy-1.10.0.dev0-cp39-cp39-macosx_12_0_arm64.whl, including the one for which I opened this issue.