multiple AccumulatedData in case of submission latency

[mtsitrin]

The AccumulatedBatchSize is cleared on submission iteration.

When m.AccumulatedBatchSize has reached the max, it will trigger toSubmit event for each subsequent block, until the submission handler iterates again.

[danwt]

In other words:
there is a loop which tracks accumulated data and submits signals whenever the accumulated mount is exceeded, but the counter is is not set to zero when submitting the signal, but rather, when consuming it, which is wrong.