Official vote tabulators at my table were very professional and efficient. However, the Michigan recount rules favor the invalidation of counted ballots but do not equally favor the inclusion of uncounted ballots (i.e., the undercount, which is the suspected problem in Michigan.) For example, certain absentee ballots (e.g., overseas military) were submitted electronically, and then transcribed by election officials onto the paper scan forms used by regular voters. These transcribed ballots were called "Duplicates" and were supposed to be numbered in order to correspond to the number of the original electronic ballot. Recount officials were required to verify that originals and duplicates corresponded, and that they were not "double-counted." However, one precinct, heavily for Clinton, had several original/duplicate pairs that were not definitively corresponding (e.g., one original ballot had no corresponding duplicate number in the forms that were scanned and thus counted.) The election official ruled that the transcriber must have forgotten to label the duplicate, but that it HAD been counted, just not labeled. However, there was no evidence to support that ruling. An alternate possibility was that the original vote had not been transcribed, and thus not counted (i.e., an undercount). Furthermore, although it was not possible at the recount to PROVE that the original HAD been counted, it might have been possible to PROVE definitively that the vote was an undercount. For example, if NONE of the counted ballots exactly matched all the marked votes on the contested ballot (included races that were not included in the recount), that would prove that this particular vote had not been counted. However, election officials insisted they were only authorized to COUNT the presidential race, and would NOT examine any ballots or portions of ballots that were not included in that narrow definition of recount. So, in all likelihood Hillary Clinton lost that one vote (out of the 200 cast) in that precinct.
A related example from another precinct comes when authorities refused to check that the sum of the individual candidate tallies exactly matches the reported vote total. Before tallying, all the votes are counted and must match the figure on the poll book. In this particular precinct, that was nearly 1100 votes. Officials counted the individual piles for only 5 candidates, and all other votes were placed in an "Other" pile. The sum of the 5 candidate tallies was 12 votes short of the total vote tally, so the "Other" pile would have to have 12 ballots if each of the 5 candidates had been correctly counted. The Other pile appeared to contain fewer than 12 ballots, which would have meant that some of the candidate tallies were incorrect. Officials refused to count that pile, even though it would have taken fewer than 10 seconds to do, citing the fact that they were not obligated to do so.
Lastly, it seemed to me that a rather large number of precincts were ineligible for recount because their seals had been broken. The counting officials asked us observers if we wished to have those precincts counted, knowing that they were not eligible for a recount. We all said no, it would be a waste of time. But I am wondering if that was a mistake (assuming the count COULD have been made). Counting would not change the election outcome (due to the recount rules), but if any of these heavily Clinton precincts had been the ones identified as statistically abnormal, we might at least have gathered evidence of fraud (or, conversely, of the absence of fraud) by counting those precincts. Surely the Stein campaign knows in which precincts those statistical aberrations occurred. If that information were conveyed to the observers, and if we were allowed to request the recount anyway, I think it would be a good idea to do so (ONLY in the statistically aberrant unrecountable precincts).
Lastly, I would suggest that the Stein recount people print out sheets with the vote tabulations from before the recount, and give these to the observers. The ballot counters do not have this information, but it is crucial to determining how vigorously we should pursue a challenge. For example, had I known that the recount "lost" Clinton 4 votes in the precinct with the absentee ballot I suspected was not counted, I would probably have contested ignoring the original Clinton vote.