Hello everyone. I have been working lately with the python bindings of the realsense2 library with the D415 model in Linux, and found something that might be a bug (or likely underdocumented error) when accumulating frames.
When I append the outputs from `pipeline.wait_for_frames()` to different python lists, it looks like it blocks the depth or color frame object and after 16 frames of appending it throws an error "Frame didn't arrive within 5000". If we try to align to the color frame, the error is "Error occured during execution of the processing block!".
However, this is easily overcome when appending a copy of said frames, which looks like a good solution and not too costly performance-wise. Before thinking of any PR to the main repo, I wanted to know whether this behavior was expected.
I am yet to understand completely the nature of the bug, but here I leave you the gists that trigger both errors, and the `copy` workaround.
If it helps, I found is not time-related but rather number of frames related, and it also is not dependent (I haven't checked all the range though) on the resolution.
Thank you very much in advance,
Pablo
https://gist.github.com/pabloalcain/65610225deb0ae9335e66e1a6c755c57
