Two identical HM-10 won't talk to each other despite flawless autoconnect

Dear fellow tinkerers, I have specific issue while working on the Arduino-Nano as telescope focuser project. I want to control stepper connected to Nano wirelessly from PC (virtual COM port, you know) running astrosoftware.
Anyway, got 2 dentical HM-10 with QUARTZ, on carrier board with 3V3 LDO, responding to AT+NAME? as "HMSoft", apparently original.
Yes it responds to MLT-BT05 command set much better than to those given in JNHuaMao datasheet, its RX is 5V-tolerant
AND autoconnection after power-off is only doable via binding (I saw they ask this question from time to time here) i.e. AT+BAND command.
So I connected consecutively each module to PC WIN10 using CP2102 USB-UART dobgle, used REALTERM, set to NO-PIN-PAIRING, AT+NOTI1 on both, connected without issues from Android [Serial Bluetooth Terminal 1.48, very good app!] and talked to modules -- sent ASCII, sent numbers, sent data file. Data exchange worked fine both ways regardless of \CR \LF \n settings in REALTERM.
Then I connected simultaneously both HM-10 to different USB ports (via CP210 USB-UART dongles), connected, AT+ROLE1, AT+INQ, AT+BAND on master, et voila, LEDs are solid.
Yet no blink of RX led when I sent something from the sibling, despite TX on sender side blinks.
Yes I can still talk normally to SLAVE from Android [if I switch off MASTER of course]-- it connects, disconnects as anticipated. But when I power MASTER -- they would pair in a second but no communication is possible between them.
Any ideas, dear fellows?

So much writing, but no mention of HOW you determined no communication was possible! Did you only discover that no communication was taking place and if so, what have you done to debug the problem?

What precisely do you want to know? Indeed, the question was "any ideas" or clues to next debug steps.

What provisions do you have to display debug messages so you can track the logic of your program?

PROGRAM?
I have no program yet, I am trying to connect one to another via REALTERM terminal soft.
I repeat, both work flawlessly when I connect to one or another (being slave at that moment, naturally) from Android BLE terminal, no issues whatsoever with data transfer.
The only issue is they fail to receive anything in either direction when one of them is master.

Which one is started first. The slave MUST be ready to receive BEFORE the master sends!

Naturally, first there is slave, then master discovers and pairs to the slave.
NEVER WORKED.
Tried 10 times.
Then tried first-master-then-slave. Same thing. I see their status (Scanning....) and connection information (connOK connLOST) but after pairing -- nothing.
I tried slave on Nano and master on WIN10, I tried over CP2102 and over original FT232 dongles -- always same thing.
Arduino or PC [slave] talks fine to mobile BLE terminal [master], but never HM-10 <-> HM-10, as seen in multiple youtube videos.

A successful connection between Blueteeth does not mean the modules are properly connected to their respective Arduinos.

Usually, when someone caught a cold he takes paracetamol or similar drug.