To turn the OT to a linear sequencer is not possible, because there are important differences between how a step sequencer and a linear sequencer work, technically speaking.
A step sequencer records events as a sequence of single descrete steps. If there are 16 steps we can use less than 16 steps but never 17 or more. We can run the sequence slow or fast, we can (on some sequencers) modify the length of single steps, we can (on some sequencers) store a step, which might be composed of more than one event (like a chord, or the p-locks on the OT), BUT we get only the maximum of steps, which the sequencer provides ... 16.
A typcial issue caused by this happens regularily, if we try to record live playing on a step sequencer.
No human plays exactly at the beat of a sequencer and so it happens that two notes may be played and fall in the same time slot of the recording step, which deletes the first note and stores the second only. The result is quite a destruction of the musical idea. This does not change in principle, if we have much more steps or run the step sequencer at higher speed.
A linear sequencer records events continuously as single events, which are composed of event-values and a time-stamp. This allows many more (all but parallel) events per time-frame to be recorded (eg. of a quarter note). Live playing is recorded almost identical to the performance, if the time-resolution is good enough and this would be above 160 ppqn (pulses per quarter note).
We can simulate a step sequencer by using a linear sequencer, but not the other way around. There is no "switch to linear" possible.
A work-around to get a higher number of events per time-frame is to use a lot of steps and run the sequence at high speed. But this will never get us near the possibilities of a real linear sequencer with a decent time-resolution.
Examples: A MPC 500, 1000, and 2500 provides 96 ppqn, a MPC 4k, 5k 960 ppqn, the new MPC Live and X should provide 960 ppqn too. That's just on another level, if we consider recording and playback with respect to time-resolution and number of events per time frame only.