Issues with SCF convergence with organometallic complexes with Group 1/2, Period 5/6 metals

[Physopholy]

Hey guys, thanks in advance (great code, by the way!).

Describe the bug
Organometallic complexes with group 1-2, period 5-6 elements seem to have convergence issues with GBSA/H2O solvation.

To Reproduce
Use xtb_input.xyz, run "xtb xtb_input.xyz --gbsa h2o --chrg 2 --opt vtight > xtb.out"

Error:

########################################################################
[ERROR] Program stopped due to fatal error
-7- Geometry optimization failed
-6- xtb_geoopt: Trying to recover from failed geometry optimization
-5- xtb_geoopt: Geometry optimization did not converge
-4- optimizer_ancopt: Could not relax structure
-3- optimizer_relax: SCF not converged, aborting...
-2- xtb_calculator_singlepoint: Electronic structure method terminated
-1- scf: Self consistent charge iterator did not converge
########################################################################

Complete printout is attached in xtb.out.

Expected behaviour
A speedy/stable relaxation, as is the norm for xTB in my experience with it so far.

Additional context

xtb version 6.6.1 (8d0f1dd) compiled by 'stahn@M-Bot' on 2023-08-01.

I/O is attached. I will say it's a bit unusual the way it is coordinating with the nitrogens (none of the other complexes strongly coordinated with the nitrogens rather than the oxygens).

It completes a few steps and then goes to the SCF step max. With the Sr-cryptand complex I did a bunch of testing (raising electronic temp., lowering to GFN 1, lowering oxidation state magnitude, removing implicit solvent, different starting structure). I managed to get it to run by slowly reintroducing the things I'd changed (e.g. adding the charge back in) - it breaks when I turn on the solvation.

I also had similar issues with a crown chelator complex with Cs/Ba (separately, not together), so a pretty similar system. Sometimes they work but have convergence issues for mRRHO evaluations (xTB-driven ofc.) in CENSO. I did not have the issue with other Group 1/2 metals. The fact that they (Cs, Ba) are so close to Sr makes me think that it isn't a coincidence.
sr_package.zip