Order reject and requote rates: what to log before blaming your edge

[LondonScalper]

The 3-in-5 rule is roughly where I ended up, though I count it per symbol and per order type. Three rejects on market orders in five minutes means something different from three on limit modifications, which some bridges bounce for reasons unrelated to liquidity.

The latency trigger I like in principle, but a jump from 50 ms to 500 ms is one you'd notice anyway because the platform starts to feel sticky. Mine is relative: two consecutive orders at more than three times my session median. That catches the milder degradation that still costs a fraction of a pip per fill.

The spam-click lockout is the most useful of the three for me, and I'd call the threshold generous. Two requests inside 1.5 seconds on one symbol is nearly always me chasing, not the venue failing. I'd also stretch the lockout past sixty seconds. A minute is barely enough to stop the hand; the reasoning behind the chase takes longer to cool.