BLE start routine freezes when the event_queue dispatching inside mbed-os-ble-utils moved to a thread

[16L-YT]

Description of defect

I'm using mbed-os-ble-utils for our BLE-enabled device. In this library the BLEProcess class has the start API to strart advertising. I tried to fork the library and make the _event_queue.dispatch_forever(); (line) inside a thread to avoid blocking the code I will execute at startup after starting the BLE advertisements .

        thread.start([this] {_event_queue.dispatch_forever();});

However the BLE stuck at

[DBG ][BLE ]: Cordio HCI reset started

I even tried to increase the priority of that thread to the highest osPriorityRealtime7 with no effect.

Note:
No extra debugging messages, even if the BLE logs are enabled:

            "cordio.trace-hci-packets": false,
            "cordio.trace-cordio-wsf-traces": true,
            "ble.trace-human-readable-enums": true,

Is there a clear reason why ?

Target(s) affected by this defect ?

AMA3B1KK-KBR-B0

Toolchain(s) (name and version) displaying this defect ?

gcc version 10.2.1 20201103 (release) (xPack GNU Arm Embedded GCC, 64-bit)

What version of Mbed-os are you using (tag or sha) ?

mbed-os-6.9.0

What version(s) of tools are you using. List all that apply (E.g. mbed-cli)

mbed-tools 7.59.0

How is this defect reproduced ?

Following the mentioned steps in the defect description.

[yahyatawil]

It seems the Bluetooth APIs are not thread-safe according to #14558.