On a system with high uptime and absurdly large dbcache and maxsigcachesize settings the time to remove conflicts is now approximately 50% of the time to process a new block.
This matters for miners time to switch to the new block.
Example:
2016-07-12 04:27:18.506662 - Load block from disk: 0.00ms [0.02s]
2016-07-12 04:27:18.506710 - Sanity checks: 0.00ms [0.01s]
2016-07-12 04:27:18.506759 - Fork checks: 0.05ms [0.03s]
2016-07-12 04:27:18.534023 - Connect 2977 transactions: 27.23ms (0.009ms/tx, 0.007ms/txin) [13.46s]
2016-07-12 04:27:18.534892 - Verify 4142 txins: 28.13ms (0.007ms/txin) [35.45s]
2016-07-12 04:27:18.546111 - Index writing: 11.21ms [4.49s]
2016-07-12 04:27:18.546149 - Callbacks: 0.05ms [0.03s]
2016-07-12 04:27:18.546589 - Connect total: 39.93ms [40.08s]
2016-07-12 04:27:18.551713 - Flush: 5.11ms [3.35s]
2016-07-12 04:27:18.551766 - Writing chainstate: 0.07ms [0.19s]
2016-07-12 04:27:18.583436 UpdateTip: new best=000000000000000002576865a40fcb68bc012366d3c08f7ae11a1df793d6eb0c height=420364 version=0x30000000 log2_work=84.974337 tx=141859971 date='2016-07-12 04:27:52' progress=1.000000 cache=770.0MiB(245590tx) warning='6 of last 100 blocks have unexpected version'
2016-07-12 04:27:18.583641 - Connect postprocess: 31.87ms [13.50s]
2016-07-12 04:27:18.583658 - Connect block: 76.99ms [57.14s]