Panphasia error encountered when using FLAMINGO descriptor

[WillJRoper]

I was investigating using MUSIC-2 to generate a test suite of zooms from FLAMINGO for SWIFT, but when I ran it I immediately encountered a Panphasia-related error:

            __    __     __  __     ______     __     ______
           /\ "-./  \   /\ \/\ \   /\  ___\   /\ \   /\  ___\
           \ \ \-./\ \  \ \ \_\ \  \ \___  \  \ \ \  \ \ \____
            \ \_\ \ \_\  \ \_____\  \/\_____\  \ \_\  \ \_____\
             \/_/  \/_/   \/_____/   \/_____/   \/_/   \/_____/

                                    this is music! version 2.0a

 Version: git rev.: 967651b, tag: N/A, branch: master
 This Release build was compiled at 13:26:16 on Sep  9 2025
 Compiled with GNU C++ version Intel(R) oneAPI DPC++/C++ Compiler 2024.2.0 (2024.2.0.20240602)
 Opening log file 'params.conf_log.txt'.
 -------------------------------------------------------------------------------
 CPU vendor string                : AMD EPYC 7542 32-Core Processor
 Available HW threads / task      : 128 (128 used)
 Total system memory (phys)       : 2051739 Mb
 Used system memory (phys)        : Max: 2338707 Mb, Min: 2338707 Mb
 Available system memory (phys)   : Max: 4294680327 Mb, Min: 4294680327 Mb
 OS/Kernel version                : Linux version 5.14 build 427
 FFTW version                     : fftw-3.3.10
 FFTW supports multi-threading    : yes
 FFTW mode                        : FFTW_ESTIMATE
 -------------------------------------------------------------------------------
 -------------------------------------------------------------------------------
 Cosmological parameters are:
  h        = 0.681           A_s      = 2.09937e-09     n_s     = 0.967
  Omega_c  = 0.256011        Omega_b  = 0.0486          Omega_m = 0.306
  Omega_r  = 7.79165e-05     Omega_nu = 0.00138906      ∑m_nu   = 0.06eV
  Omega_DE = 0.693922        w_0      = -1              w_a     = 0
Warning :  Radiation enabled, using Omega_r=7.79165e-05 internally for backscaling.
Warning :  Make sure your sim code supports this, otherwise set [cosmology] / ZeroRadiation=true.
 Linear growth factors: D+_target = 1, D+_start = 0.0401361
 -------------------------------------------------------------------------------
 - Selecting transfer function plug-in 'CLASS'...
 CLASS: Using A_s=2.09937e-09 to normalise the transfer function.
 Computing transfer function via ClassEngine...
 CLASS took 5.1774 s.
 CLASS table contains k = 1.03649e-05 to 168.1 h Mpc-1.
 Measured sigma_8 for given PS normalisation is 0.0323797
  . TF supports distinct CDM+baryons : yes
  . TF maximum wave number        : 168.1 h/Mpc
 - starting at a=0.03125
 - Selecting region generator plug-in 'ellipsoid'...
 Region center for ellipsoid determined at
         xc = ( 0.565489 0.516854 0.612741 )
 Ellipsoid matrix determined as
              / 2022.37  -727.432 647.419  \
          A = | -727.432 1231.42  156.455  |
              \ 647.419  156.455  1635.12  /
 - Selecting random number generator plug-in 'PANPHASIA'...
 [Panph6,
 Error reading panphasian descriptor. Error number:           1

Here is the config file I was testing:

[setup]
boxlength    = 135.54
zstart       = 31
levelmin     = 7
levelmin_TF  = 9
levelmax     = 11
baryons      = no
use_2LPT     = no
use_LLA      = no
padding      = 8

# Ellipsoidal refinement region defined on unity cube
# This minimum bounding ellipsoid has been obtained from
# particles within 2 R200c of the halo center
region                     = ellipsoid
region_ellipsoid_matrix[0] = 2022.3683, -727.43207, 647.41876
region_ellipsoid_matrix[1] = -727.43207, 1231.4218, 156.45541
region_ellipsoid_matrix[2] = 647.41876, 156.45541, 1635.1184
region_ellipsoid_center    = 0.56548864, 0.5168543, 0.6127409

[cosmology]
Omega_m         = 0.306
Omega_b         = 0.0486
Omega_L         = 0.693922
H0              = 68.1
nspec           = 0.967
A_s             = 2.0993736148e-09
Tcmb            = 2.7255
k_p             = 0.05
N_ur            = 2.0308
m_nu1           = 0.06
m_nu2           = 0
m_nu3           = 0
deg_nu1         = 1.0
deg_nu2         = 0.0
deg_nu3         = 0.0
w_0             = -1.0
w_a             = 0.0

ZeroRadiation   = false

# Transfer function
transfer        = CLASS
ztarget         = 0.0

[random]
generator      = PANPHASIA
descriptor     = [Panph6,L18,(20057,13876,16144),S1,KK1025,CH-999,FLAMINGO200]

[poisson]
fft_fine      = yes
accuracy      = 1e-6
grad_order    = 6
laplace_order = 6

[output]
format                  = swift
filename                = ics_swift.hdf5

I have tested the Auriga configuration file at data/ExampleConfigs/AURIGA6_ics.conf, and this runs fine, but there are some clear differences between the descriptors with a different number of elements. I can't claim to intimately understand Panphasia, but removing the KK element and swapping to "Panph1" then ran without issue (but, of course, that is no longer the descriptor for the ICs I want to generate zooms from).

Is this a fundamental limitation or something I've misunderstood?