So this may be the most bizarre problem I’ve ever seen: I pulled a copy of firmware 2.0.7 source code, re-compiled it, and uploaded, and the problem is gone! From the last time 2.0.7 firmware is generated, the only thing I can think of is I upgraded my Linux system, so there is non-zero possibility that this is a compiler bug. Here are my observations:
1) I can confirm that firmware 2.0.7 we pre-flahsed on all controllers and checked into Github all have the reported issue.
2) The newly compiled firmware 2.0.7 does NOT have this issue.
3) The newly compiled 2.0.7 and the original 2.0.7 have exactly the same size (55612 binary code size) but the content is clearly different.
No. 3 above is quite surprising. In any case, I am not able to debug the issue (unless I can find the avr-gcc version that I used to compile the original 2.0.7). For now the solution is to either upgrade to 2.0.8, or if you want to stick with 2.0.7, I can post the newly compiled 2.0.7.