TDA8932 has problems with 4S BMS

I use one of those tiny TDA8932 boards to amp an fr-br-box.
The amp sits inside the cabinet together with analog preamp i built around an TL074.
That preamp offers a mono-summer and tone controls.
Everything worked like a charme .... but then i had an idea....
Yesterday i built a battery-holder with four 18650 cells into the cabinet.
I glued an bms-pcb to the backside of the battery-holder and connected everything as shown in the pic.
LOAD are the amp and preamp in parallel, CHARGER is the DC-jack that i already had at the backplate of the cabinet, only thing that is not shown is a switch that (dis)connects the LOAD from the plus-line.
I attached precharged cells and measured 16.5V.
As soon as the LOAD is connected the voltage drops to about 300mV.
That's still the case, when the preamp is disconnected.
And now the crazyness starts:
I plug 18V DC to the CHARGER inputs.
The amps starts playing (quite noisy).
I disconnect the DC plug.
The amp keeps playing from the batteries - now very nicely, because almost noise free.

I don't accept the emergency-start with the DC plug as a solution.
Please help me explain the nature of the problem.
Please help me find a better solution.
Thanks for all replies!

Are the cells the correct chemistry for the BMS?
The BMS board says it's made for Li-Polymer.
My cells are MOLI IBR-18650B, the internet says they are Li-Ion.
I thought that the combination would be ok, is it?

Is the 300mV across the load or across the 4 cells?

300mV across the load.

Some sort of latch up when on battery power ?

My english is not good enough to understand what you mean by 'latch up'.
There is a huge driver excursion when i connect the dc-plug.