Occasional scrambling of data on Bluetooth UART after automatic reconnects?

I think, I have made an interesting observation:

I'm greping now for XON, XOFF, UART and 'characteristic write' in adb logcat.

When everything is working fine, I can see many lines of 'characteristic write' writing to the UART. In between that, I see some flow control messages sending XOFF and then XON quite quickly after that.

So far, the timeout issue has reproduced itself only once while adb was connected. In that case, I saw XOFF being sent as one of the last messages with no XON following. The lines 'characteristic write' stop after that, the Bangle goes to timeout, but the http requests are still handled and logged by gadgetbridge.

I'm wondering whether the issue is that in case that a bluetooth connection loss falls exactly into the time between XOFF and XON, the XON gets lost and this is what freezes the communication gadgetbridge -> bangle. The XON is possibly never repeated. The larger the data volume, the more flow control kicks in and the more likely it is to run into this issue. Maybe a possible solution would be to send XON automatically on every fresh bluetooth (re-)connect?

Another observation: I'm printing the exception message of Bangle.http(...) to the display and when the Bluetooth goes out of range, I always get a "Timeout" exception first, which changes after a few seconds to "Bluetooth not connected". So it seems that the Bangle needs some seconds to understand that the Bluetooth connection has been fully lost, it doesn't realize it instantaneously. I'm wondering whether this could be the moment, during which XON gets lost.