It feels like a pretty typical groupbyfunction example:

cols = ['流水号', '处理人', '处理时间']
data = [[10000, '张三', '2016-10-01'],
        [10000, '李四', '2016-10-02'],
        [10001, '王五', '2016-10-01'],
        [10002, '赵六', '2016-10-03'],
        [10001, '黄七', '2016-10-02'],
        [10000, '吴八', '2016-10-03']]
frame = pd.DataFrame(data,columns=cols)

def combination(names):
    return ','.join(names)
    
frame.groupby('流水号').aggregate(combination)

The output should be like this:

In [12]: frame.groupby('流水号').aggregate(combinition)
Out[12]: 
            处理人                              处理时间
流水号                                              
10000  张三,李四,吴八  2016-10-01,2016-10-02,2016-10-03
10001     王五,黄七             2016-10-01,2016-10-02
10002        赵六  

python3

import pandas as pd

cols = ['流水号', '处理人', '处理时间']
data = [[10000, '张三', '2016-10-01'],
        [10000, '李四', '2016-10-02'],
        [10001, '王五', '2016-10-01'],
        [10002, '赵六', '2016-10-03'],
        [10001, '黄七', '2016-10-02'],
        [10000, '吴八', '2016-10-03']]

df = pd.DataFrame(data,columns=cols)
grp = [(n, ','.join([r for r in set(df[df['流水号']==n]['处理人'])]))
                       for n in set(df['流水号'])]

df2 = pd.DataFrame(grp, columns=cols[:-1])
print(df)
print(df2)