I am running a software timer with a 1 tick period which is started from an ISR (using xTimerStartFromISR) and then stops itself (using xTimerStop).

After the call to xTimerStop returns pdPASS the timer will sometimes continue running. This comes as a surprise to me as the timer.c code looks like it processes all the commands in the queue between invocations of the timer handler.

My current workaround is to use a local flag to indicate if the timer should be stopped and try calling xTimerStop again.

Forgot to mention this is FreeRTOS version 7.4.0

When you send the stop command the command is posted to a queue that is drained by the timer service task, so the return value indicates the command was sent to the queue not that the command has been processed. When the command is processed depends on the priority of the timer service task. If the timer service task has the highest priority of all your tasks then it will get processed right away.

I'm not sure it is viable to have a software timer that has the same frequency as the tick. The tick sets the time base, so if your software timer time period is exactly the system time base then it will need to run all the time. Maybe it is preventing the command being processed?

I use timers as Task alternatives for most modules in my system as they:
1) share a stack, saving a significant amount of RAM.
2) run cooperatively, simplifying shared data access

This particular 'timer' is effectively a block of code which runs on an event triggered by an ISR. I *could* have done it with a task, but it would be sitting there 99.99% of the time hogging stack space and doing nothing.

Certainly I am aware that this is not what timers were designed for, but I expected that the stop command (called from the timer handler) would be processed before the timer handler was invoked again.