There is none.
in add_waiting_time we do
waiting_times[category].put(halt->get_basis_pos(), ....)
But when halt is deleted, its basis_pos is still here.
Edit:
Should be fixed in my waiting_times_fix (https://github.com/inkelyad/simutrans-experimental/commits/waiting_times_fix) branch.
Windows binaries (https://github.com/downloads/inkelyad/simutrans-experimental/waiting_times_fix_a973c9.7z) of my waiting_time_fix branch.
Is this patch intended to address the issue reported in this (http://forum.simutrans.com/index.php?topic=6210.msg0#new) thread?
Edit: It seems that it is indeed intended to be such a fix, looking at the code. But it looks as if it might be too aggressive a fix: it appends a waiting time of 9 minutes to the end of the fixed-sized list of 16 of the last waiting times every 256 steps, which is every real life minute or so per stop running at normal speed. Also, this will have adverse effects if the actual average waiting time of a stop is below nine minutes: there is no reason in that case to increase average waiting times.
Edit 2: A revised version of this has now been pushed to the 9.x branch - I'd be grateful for any testing.
Incorporated into 9.0.